@article{banerjee2023learning,
abstract = {Room-temperature semiconductor radiation detectors (RTSD) have broad applications in medical imaging, homeland security, astrophysics and others. RTSDs such as CdZnTe, CdTe are often pixelated, and characterization of these detectors at micron level can benefit 3-D event reconstruction at sub-pixel level. Material defects alongwith electron and hole charge transport properties need to be characterized which requires several experimental setups and is labor intensive. The current state-of-art approaches characterize each detector pixel, considering the detector in bulk. In this article, we propose a new microscopic learning-based physical models of RTSD based on limited data compared to what is dictated by the physical equations. Our learning models uses a physical charge transport considering trapping centers. Our models learn these material properties in an indirect manner from the measurable signals at the electrodes and/or free and/or trapped charges distributed in the RTSD for electron–hole charge pair injections in the material. Based on the amount of data used during training our physical model, our algorithm characterizes the detector for charge drifts, trapping, detrapping and recombination coefficients considering multiple trapping centers or as a single equivalent trapping center. The RTSD is segmented into voxels spatially, and in each voxel, the material properties are modeled as learnable parameters. Depending on the amount of data, our models can characterize the RTSD either completely or in an equivalent manner.},
author = {Banerjee, Srutarshi and Rodrigues, Miesher and Ballester, Manuel and Vija, Alexander Hans and Katsaggelos, Aggelos K.},
doi = {10.1038/s41598-022-27125-7},
issn = {20452322},
journal = {Scientific Reports},
number = {1},
pages = {168},
pmid = {36599876},
publisher = {Nature Publishing Group UK London},
title = {{Learning-based physical models of room-temperature semiconductor detectors with reduced data}},
volume = {13},
year = {2023}
}

Room-temperature semiconductor radiation detectors (RTSD) have broad applications in medical imaging, homeland security, astrophysics and others. RTSDs such as CdZnTe, CdTe are often pixelated, and characterization of these detectors at micron level can benefit 3-D event reconstruction at sub-pixel level. Material defects alongwith electron and hole charge transport properties need to be characterized which requires several experimental setups and is labor intensive. The current state-of-art approaches characterize each detector pixel, considering the detector in bulk. In this article, we propose a new microscopic learning-based physical models of RTSD based on limited data compared to what is dictated by the physical equations. Our learning models uses a physical charge transport considering trapping centers. Our models learn these material properties in an indirect manner from the measurable signals at the electrodes and/or free and/or trapped charges distributed in the RTSD for electron–hole charge pair injections in the material. Based on the amount of data used during training our physical model, our algorithm characterizes the detector for charge drifts, trapping, detrapping and recombination coefficients considering multiple trapping centers or as a single equivalent trapping center. The RTSD is segmented into voxels spatially, and in each voxel, the material properties are modeled as learnable parameters. Depending on the amount of data, our models can characterize the RTSD either completely or in an equivalent manner.

@article{ruiz2023probabilistic,
author = {Ruiz, Pablo and Morales-{\'{A}}lvarez, Pablo and Coughlin, Scott and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.knosys.2022.110183},
issn = {09507051},
journal = {Knowledge-Based Systems},
pages = {110183},
publisher = {Elsevier},
title = {{Probabilistic fusion of crowds and experts for the search of gravitational waves}},
volume = {261},
year = {2023}
}

@article{lopez2023deep,
author = {Lopez-Perez, Miguel and Morales-Alvarez, Pablo and Cooper, Lee A. D. and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/access.2023.3237990},
issn = {21693536},
journal = {IEEE Access},
pages = {6922--6934},
publisher = {IEEE},
title = {{Deep Gaussian Processes for Classification With Multiple Noisy Annotators. Application to Breast Cancer Tissue Classification}},
volume = {11},
year = {2023}
}

@article{Yunan2023,
author = {Wu, Yunan and Dravid, Amil and Wehbe, Ramsey Michael and Katsaggelos, Aggelos K},
doi = {https://doi.org/10.48550/arXiv.2301.08798},
journal = {arXiv preprint arXiv:2301.08798},
title = {{DeepCOVID-Fuse: A Multi-modality Deep Learning Model Fusing Chest X-Radiographs and Clinical Variables to Predict COVID-19 Risk Levels}},
year = {2023}
}

@article{Fragos2023,
abstract = {Most massive stars are members of a binary or a higher-order stellar system, where the presence of a binary companion can decisively alter their evolution via binary interactions. Interacting binaries are also important astrophysical laboratories for the study of compact objects. Binary population synthesis studies have been used extensively over the last two decades to interpret observations of compact-object binaries and to decipher the physical processes that lead to their formation. Here, we present POSYDON , a novel, publicly available, binary population synthesis code that incorporates full stellar structure and binary-evolution modeling, using the MESA code, throughout the whole evolution of the binaries. The use of POSYDON enables the self-consistent treatment of physical processes in stellar and binary evolution, including: realistic mass-transfer calculations and assessment of stability, internal angular-momentum transport and tides, stellar core sizes, mass-transfer rates, and orbital periods. This paper describes the detailed methodology and implementation of POSYDON , including the assumed physics of stellar and binary evolution, the extensive grids of detailed single- and binary-star models, the postprocessing, classification, and interpolation methods we developed for use with the grids, and the treatment of evolutionary phases that are not based on precalculated grids. The first version of POSYDON targets binaries with massive primary stars (potential progenitors of neutron stars or black holes) at solar metallicity.},
archivePrefix = {arXiv},
arxivId = {2202.05892},
author = {Fragos, Tassos and {J. Andrews}, Jeff and Bavera, Simone S. and Berry, Christopher P. L. and Coughlin, Scott and Dotter, Aaron and Giri, Prabin and Kalogera, Vicky and Katsaggelos, Aggelos and Kovlakas, Konstantinos and Lalvani, Shamal and Misra, Devina and Srivastava, Philipp M. and Qin, Ying and Rocha, Kyle A. and Rom{\'{a}}n-Garza, Jaime and Serra, Juan Gabriel and Stahle, Petter and Sun, Meng and Teng, Xu and Trajcevski, Goce and Tran, Nam Hai and Xing, Zepei and Zapartas, Emmanouil and Zevin, Michael},
chapter = {45},
doi = {10.3847/1538-4365/ac90c1},
eprint = {2202.05892},
isbn = {0067-0049 1538-4365},
issn = {0067-0049},
journal = {The Astrophysical Journal Supplement Series},
month = {feb},
number = {2},
pages = {45},
title = {{POSYDON: A General-purpose Population Synthesis Code with Detailed Binary-evolution Simulations}},
url = {http://arxiv.org/abs/2202.05892 https://iopscience.iop.org/article/10.3847/1538-4365/ac90c1},
volume = {264},
year = {2023}
}

Most massive stars are members of a binary or a higher-order stellar system, where the presence of a binary companion can decisively alter their evolution via binary interactions. Interacting binaries are also important astrophysical laboratories for the study of compact objects. Binary population synthesis studies have been used extensively over the last two decades to interpret observations of compact-object binaries and to decipher the physical processes that lead to their formation. Here, we present POSYDON , a novel, publicly available, binary population synthesis code that incorporates full stellar structure and binary-evolution modeling, using the MESA code, throughout the whole evolution of the binaries. The use of POSYDON enables the self-consistent treatment of physical processes in stellar and binary evolution, including: realistic mass-transfer calculations and assessment of stability, internal angular-momentum transport and tides, stellar core sizes, mass-transfer rates, and orbital periods. This paper describes the detailed methodology and implementation of POSYDON , including the assumed physics of stellar and binary evolution, the extensive grids of detailed single- and binary-star models, the postprocessing, classification, and interpolation methods we developed for use with the grids, and the treatment of evolutionary phases that are not based on precalculated grids. The first version of POSYDON targets binaries with massive primary stars (potential progenitors of neutron stars or black holes) at solar metallicity.

@article{Jane2022,
abstract = {Understanding the noise in gravitational-wave detectors is central to detecting and interpreting gravitational-wave signals. Glitches are transient, non-Gaussian noise features that can have a range of environmental and instrumental origins. The Gravity Spy project uses a machine-learning algorithm to classify glitches based upon their time–frequency morphology. The resulting set of classified glitches can be used as input to detector-characterisation investigations of how to mitigate glitches, or data-analysis studies of how to ameliorate the impact of glitches. Here we present the results of the Gravity Spy analysis of data up to the end of the third observing run of advanced laser interferometric gravitational-wave observatory (LIGO). We classify 233981 glitches from LIGO Hanford and 379805 glitches from LIGO Livingston into morphological classes. We find that the distribution of glitches differs between the two LIGO sites. This highlights the potential need for studies of data quality to be individually tailored to each gravitational-wave observatory.},
archivePrefix = {arXiv},
arxivId = {2208.12849},
author = {Glanzer, Jane and Banagiri, S and Coughlin, S B and Soni, Siddharth and Zevin, Michael and Berry, Christopher Philip Luke and Patane, Oli and Bahaadini, Sara and Rohani, Neda and Crowston, Kevin and Kalogera, V and {\O}sterlund, Carsten and Trouille, Laura and Katsaggelos, A},
doi = {10.1088/1361-6382/acb633},
eprint = {2208.12849},
issn = {0264-9381},
journal = {Classical and Quantum Gravity},
month = {mar},
number = {6},
pages = {065004},
title = {{Data quality up to the third observing run of advanced LIGO: Gravity Spy glitch classifications}},
url = {https://iopscience.iop.org/article/10.1088/1361-6382/acb633},
volume = {40},
year = {2023}
}

Understanding the noise in gravitational-wave detectors is central to detecting and interpreting gravitational-wave signals. Glitches are transient, non-Gaussian noise features that can have a range of environmental and instrumental origins. The Gravity Spy project uses a machine-learning algorithm to classify glitches based upon their time–frequency morphology. The resulting set of classified glitches can be used as input to detector-characterisation investigations of how to mitigate glitches, or data-analysis studies of how to ameliorate the impact of glitches. Here we present the results of the Gravity Spy analysis of data up to the end of the third observing run of advanced laser interferometric gravitational-wave observatory (LIGO). We classify 233981 glitches from LIGO Hanford and 379805 glitches from LIGO Livingston into morphological classes. We find that the distribution of glitches differs between the two LIGO sites. This highlights the potential need for studies of data quality to be individually tailored to each gravitational-wave observatory.

@article{Zixiao2023,
author = {Yu, Zixiao and Wang, Haohong and Katsaggelos, Aggelos K and Ren, Jian},
journal = {IEEE Internet of Things Journal},
title = {{A Novel Automatic Content Generation And Optimization Framework}},
year = {2023}
}

@article{Henry2022,
abstract = {Macro x-ray fluorescence (XRF) imaging of cultural heritage objects, while a popular non-invasive technique for providing elemental distribution maps, is a slow acquisition process in acquiring high signal-to-noise ratio XRF volumes. Typically on the order of tenths of a second per pixel, a raster scanning probe counts the number of photons at different energies emitted by the object under x-ray illumination. In an effort to reduce the scan times without sacrificing elemental map and XRF volume quality, we propose using dictionary learning with a Poisson noise model as well as a color image-based prior to restore noisy, rapidly acquired XRF data.},
archivePrefix = {arXiv},
arxivId = {2206.01740},
author = {Chopp, Henry and McGeachy, Alicia and Alfeld, Matthias and Cossairt, Oliver and Walton, Marc and Katsaggelos, Aggelos},
eprint = {2206.01740},
journal = {arXiv preprint arXiv:2206.01740},
month = {jun},
title = {{Denoising Fast X-Ray Fluorescence Raster Scans of Paintings}},
url = {http://arxiv.org/abs/2206.01740},
year = {2022}
}

Macro x-ray fluorescence (XRF) imaging of cultural heritage objects, while a popular non-invasive technique for providing elemental distribution maps, is a slow acquisition process in acquiring high signal-to-noise ratio XRF volumes. Typically on the order of tenths of a second per pixel, a raster scanning probe counts the number of photons at different energies emitted by the object under x-ray illumination. In an effort to reduce the scan times without sacrificing elemental map and XRF volume quality, we propose using dictionary learning with a Poisson noise model as well as a color image-based prior to restore noisy, rapidly acquired XRF data.

@article{Sara2022,
abstract = {In this paper, leveraging the capabilities of neural networks for modeling the non-linearities that exist in the data, we propose several models that can project data into a low dimensional, discriminative, and smooth manifold. The proposed models can transfer knowledge from the domain of known classes to a new domain where the classes are unknown. A clustering algorithm is further applied in the new domain to find potentially new classes from the pool of unlabeled data. The research problem and data for this paper originated from the Gravity Spy project which is a side project of Advanced Laser Interferometer Gravitational-wave Observatory (LIGO). The LIGO project aims at detecting cosmic gravitational waves using huge detectors. However non-cosmic, non-Gaussian disturbances known as "glitches", show up in gravitational-wave data of LIGO. This is undesirable as it creates problems for the gravitational wave detection process. Gravity Spy aids in glitch identification with the purpose of understanding their origin. Since new types of glitches appear over time, one of the objective of Gravity Spy is to create new glitch classes. Towards this task, we offer a methodology in this paper to accomplish this.},
archivePrefix = {arXiv},
arxivId = {2205.13672},
author = {Bahaadini, Sara and Wu, Yunan and Coughlin, Scott and Zevin, Michael and Katsaggelos, Aggelos K.},
eprint = {2205.13672},
journal = {arXiv e-prints},
month = {may},
title = {{Discriminative Dimensionality Reduction using Deep Neural Networks for Clustering of LIGO Data}},
url = {http://arxiv.org/abs/2205.13672},
year = {2022}
}

In this paper, leveraging the capabilities of neural networks for modeling the non-linearities that exist in the data, we propose several models that can project data into a low dimensional, discriminative, and smooth manifold. The proposed models can transfer knowledge from the domain of known classes to a new domain where the classes are unknown. A clustering algorithm is further applied in the new domain to find potentially new classes from the pool of unlabeled data. The research problem and data for this paper originated from the Gravity Spy project which is a side project of Advanced Laser Interferometer Gravitational-wave Observatory (LIGO). The LIGO project aims at detecting cosmic gravitational waves using huge detectors. However non-cosmic, non-Gaussian disturbances known as "glitches", show up in gravitational-wave data of LIGO. This is undesirable as it creates problems for the gravitational wave detection process. Gravity Spy aids in glitch identification with the purpose of understanding their origin. Since new types of glitches appear over time, one of the objective of Gravity Spy is to create new glitch classes. Towards this task, we offer a methodology in this paper to accomplish this.

@article{Semih2022,
abstract = {Ptychography is a well-established coherent diffraction imaging technique that enables non-invasive imaging of samples at a nanometer scale. It has been extensively used in various areas such as the defense industry or materials science. One major limitation of ptychography is the long data acquisition time due to mechanical scanning of the sample; therefore, approaches to reduce the scan points are highly desired. However, reconstructions with less number of scan points lead to imaging artifacts and significant distortions, hindering a quantitative evaluation of the results. To address this bottleneck, we propose a generative model combining deep image priors with deep generative priors. The self-training approach optimizes the deep generative neural network to create a solution for a given dataset. We complement our approach with a prior acquired from a previously trained discriminator network to avoid a possible divergence from the desired output caused by the noise in the measurements. We also suggest using the total variation as a complementary before combat artifacts due to measurement noise. We analyze our approach with numerical experiments through different probe overlap percentages and varying noise levels. We also demonstrate improved reconstruction accuracy compared to the state-of-the-art method and discuss the advantages and disadvantages of our approach.},
archivePrefix = {arXiv},
arxivId = {2205.02397},
author = {Barutcu, Semih and G{\"{u}}rsoy, Doğa and Katsaggelos, Aggelos K.},
eprint = {2205.02397},
journal = {arXiv preprint arXiv:2205.02397},
month = {may},
title = {{Compressive Ptychography using Deep Image and Generative Priors}},
url = {http://arxiv.org/abs/2205.02397},
year = {2022}
}

Ptychography is a well-established coherent diffraction imaging technique that enables non-invasive imaging of samples at a nanometer scale. It has been extensively used in various areas such as the defense industry or materials science. One major limitation of ptychography is the long data acquisition time due to mechanical scanning of the sample; therefore, approaches to reduce the scan points are highly desired. However, reconstructions with less number of scan points lead to imaging artifacts and significant distortions, hindering a quantitative evaluation of the results. To address this bottleneck, we propose a generative model combining deep image priors with deep generative priors. The self-training approach optimizes the deep generative neural network to create a solution for a given dataset. We complement our approach with a prior acquired from a previously trained discriminator network to avoid a possible divergence from the desired output caused by the noise in the measurements. We also suggest using the total variation as a complementary before combat artifacts due to measurement noise. We analyze our approach with numerical experiments through different probe overlap percentages and varying noise levels. We also demonstrate improved reconstruction accuracy compared to the state-of-the-art method and discuss the advantages and disadvantages of our approach.

@article{banerjee2022joint,
abstract = {We present a novel adaptive multimodal intensity-event algorithm to optimize an overall objective of object tracking under bit rate constraints for a host-chip architecture. The chip is a computationally resource-constrained device acquiring high-resolution intensity frames and events, while the host is capable of performing computationally expensive tasks. We develop a joint intensity-neuromorphic event rate-distortion compression framework with a quadtree (QT)-based compression of intensity and events scheme. The goal of this compression framework is to optimally allocate bits to the intensity frames and neuromorphic events based on the minimum distortion at a given communication channel capacity. The data acquisition on the chip is driven by the presence of objects of interest in the scene as detected by an object detector. The most informative intensity and event data are communicated to the host under rate constraints so that the best possible tracking performance is obtained. The detection and tracking of objects in the scene are done on the distorted data at the host. Intensity and events are jointly used in a fusion framework to enhance the quality of the distorted images, in order to improve the object detection and tracking performance. The performance assessment of the overall system is done in terms of the multiple object tracking accuracy (MOTA) score. Compared with using intensity modality only, there is an improvement in MOTA using both these modalities in different scenarios.},
author = {Banerjee, Srutarshi and Chopp, Henry H. and Zhang, Jianping and Wang, Zihao W. and Kang, Peng and Cossairt, Oliver and Katsaggelos, Aggelos},
doi = {10.1109/TNNLS.2022.3214779},
issn = {21622388},
journal = {IEEE Transactions on Neural Networks and Learning Systems},
keywords = {Compressed domain object detection and tracking,Computer architecture,Image reconstruction,Neuromorphics,Object detection,Optimization,Rate-distortion,Task analysis,dynamic programming (DP),joint intensity-event imaging system,joint intensity-event rate-distortion optimization,quadtree (QT) segmentation},
publisher = {IEEE},
title = {{A Joint Intensity-Neuromorphic Event Imaging System With Bandwidth-Limited Communication Channel}},
year = {2022}
}

We present a novel adaptive multimodal intensity-event algorithm to optimize an overall objective of object tracking under bit rate constraints for a host-chip architecture. The chip is a computationally resource-constrained device acquiring high-resolution intensity frames and events, while the host is capable of performing computationally expensive tasks. We develop a joint intensity-neuromorphic event rate-distortion compression framework with a quadtree (QT)-based compression of intensity and events scheme. The goal of this compression framework is to optimally allocate bits to the intensity frames and neuromorphic events based on the minimum distortion at a given communication channel capacity. The data acquisition on the chip is driven by the presence of objects of interest in the scene as detected by an object detector. The most informative intensity and event data are communicated to the host under rate constraints so that the best possible tracking performance is obtained. The detection and tracking of objects in the scene are done on the distorted data at the host. Intensity and events are jointly used in a fusion framework to enhance the quality of the distorted images, in order to improve the object detection and tracking performance. The performance assessment of the overall system is done in terms of the multiple object tracking accuracy (MOTA) score. Compared with using intensity modality only, there is an improvement in MOTA using both these modalities in different scenarios.

@article{Semih2022a,
abstract = {Ptychography is a well-studied phase imaging method that makes non-invasive imaging possible at a nanometer scale. It has developed into a mainstream technique with various applications across a range of areas such as material science or the defense industry. One major drawback of ptychography is the long data acquisition time due to the high overlap requirement between adjacent illumination areas to achieve a reasonable reconstruction. Traditional approaches with reduced overlap between scanning areas result in reconstructions with artifacts. In this paper, we propose complementing sparsely acquired or undersampled data with data sampled from a deep generative network to satisfy the oversampling requirement in ptychography. Because the deep generative network is pre-trained and its output can be computed as we collect data, the experimental data and the time to acquire the data can be reduced. We validate the method by presenting the reconstruction quality compared to the previously proposed and traditional approaches and comment on the strengths and drawbacks of the proposed approach.},
archivePrefix = {arXiv},
arxivId = {2207.14392},
author = {Barutcu, Semih and Katsaggelos, Aggelos K. and G{\"{u}}rsoy, Doğa},
eprint = {2207.14392},
journal = {arXiv preprint arXiv:2207.14392},
month = {jul},
title = {{A Deep Generative Approach to Oversampling in Ptychography}},
url = {http://arxiv.org/abs/2207.14392},
year = {2022}
}

Ptychography is a well-studied phase imaging method that makes non-invasive imaging possible at a nanometer scale. It has developed into a mainstream technique with various applications across a range of areas such as material science or the defense industry. One major drawback of ptychography is the long data acquisition time due to the high overlap requirement between adjacent illumination areas to achieve a reasonable reconstruction. Traditional approaches with reduced overlap between scanning areas result in reconstructions with artifacts. In this paper, we propose complementing sparsely acquired or undersampled data with data sampled from a deep generative network to satisfy the oversampling requirement in ptychography. Because the deep generative network is pre-trained and its output can be computed as we collect data, the experimental data and the time to acquire the data can be reduced. We validate the method by presenting the reconstruction quality compared to the previously proposed and traditional approaches and comment on the strengths and drawbacks of the proposed approach.

@article{bari2022psychotic,
abstract = {Background: The COVID-19 disease results from infection by the SARS-CoV-2 virus to produce a range of mild to severe physical, neurological, and mental health symptoms. The COVID-19 pandemic has indirectly caused significant emotional distress, triggering the emergence of mental health symptoms in individuals who were not previously affected or exacerbating symptoms in those with existing mental health conditions. Emotional distress and certain mental health conditions can lead to violent ideation and disruptive behavior, including aggression, threatening acts, deliberate harm toward other people or animals, and inattention to or noncompliance with education or workplace rules. Of the many mental health conditions that can be associated with violent ideation and disruptive behavior, psychosis can evidence greater vulnerability to unpredictable changes and being at a greater risk for them. Individuals with psychosis can also be more susceptible to contracting COVID-19 disease. Objective: This study aimed to investigate whether violent ideation, disruptive behavior, or psychotic symptoms were more prevalent in a population with COVID-19 and did not precede the pandemic. Methods: In this preliminary study, we analyzed questionnaire responses from a population sample (N=366), received between the end of February 2021 and the start of March 2021 (1 year into the COVID-19 pandemic), regarding COVID-19 illness, violent ideation, disruptive behavior, and psychotic symptoms. Using the Wilcoxon rank sum test followed by multiple comparisons correction, we compared the self-reported frequency of these variables for 3 time windows related to the past 1 month, past 1 month to 1 year, and >1 year ago among the distributions of people who answered whether they tested positive or were diagnosed with COVID-19 by a clinician. We also used multivariable logistic regression with iterative resampling to investigate the relationship between these variables occurring >1 year ago (ie, before the pandemic) and the likelihood of contracting COVID-19. Results: We observed a significantly higher frequency of self-reported violent ideation, disruptive behavior, and psychotic symptoms, for all 3 time windows of people who tested positive or were diagnosed with COVID-19 by a clinician. Using multivariable logistic regression, we observed 72% to 94% model accuracy for an increased incidence of COVID-19 in participants who reported violent ideation, disruptive behavior, or psychotic symptoms >1 year ago. Conclusions: This preliminary study found that people who reported a test or clinician diagnosis of COVID-19 also reported higher frequencies of violent ideation, disruptive behavior, or psychotic symptoms across multiple time windows, indicating that they were not likely to be the result of COVID-19. In parallel, participants who reported these behaviors >1 year ago (ie, before the pandemic) were more likely to be diagnosed with COVID-19, suggesting that violent ideation, disruptive behavior, in addition to psychotic symptoms, were associated with COVID-19 with an approximately 70% to 90% likelihood.},
author = {Bari, Sumra and Vike, Nicole L. and Stetsiv, Khrystyna and Woodward, Sean and Lalvani, Shamal and Stefanopoulos, Leandros and Kim, Byoung Woo and Maglaveras, Nicos and Breiter, Hans C. and Katsaggelos, Aggelos K.},
doi = {10.2196/36444},
issn = {2561-326X},
journal = {JMIR Formative Research},
keywords = {COVID-19,delusions,disruptive behavior,distress,machine learning,mental health,pandemic,paranoia,psychological health,psychosis,psychotic symptoms,risk,stress,violent ideation},
month = {aug},
number = {8},
pages = {e36444},
publisher = {JMIR Publications},
title = {{The Prevalence of Psychotic Symptoms, Violent Ideation, and Disruptive Behavior in a Population With SARS-CoV-2 Infection: Preliminary Study}},
url = {https://formative.jmir.org/2022/8/e36444},
volume = {6},
year = {2022}
}

Background: The COVID-19 disease results from infection by the SARS-CoV-2 virus to produce a range of mild to severe physical, neurological, and mental health symptoms. The COVID-19 pandemic has indirectly caused significant emotional distress, triggering the emergence of mental health symptoms in individuals who were not previously affected or exacerbating symptoms in those with existing mental health conditions. Emotional distress and certain mental health conditions can lead to violent ideation and disruptive behavior, including aggression, threatening acts, deliberate harm toward other people or animals, and inattention to or noncompliance with education or workplace rules. Of the many mental health conditions that can be associated with violent ideation and disruptive behavior, psychosis can evidence greater vulnerability to unpredictable changes and being at a greater risk for them. Individuals with psychosis can also be more susceptible to contracting COVID-19 disease. Objective: This study aimed to investigate whether violent ideation, disruptive behavior, or psychotic symptoms were more prevalent in a population with COVID-19 and did not precede the pandemic. Methods: In this preliminary study, we analyzed questionnaire responses from a population sample (N=366), received between the end of February 2021 and the start of March 2021 (1 year into the COVID-19 pandemic), regarding COVID-19 illness, violent ideation, disruptive behavior, and psychotic symptoms. Using the Wilcoxon rank sum test followed by multiple comparisons correction, we compared the self-reported frequency of these variables for 3 time windows related to the past 1 month, past 1 month to 1 year, and >1 year ago among the distributions of people who answered whether they tested positive or were diagnosed with COVID-19 by a clinician. We also used multivariable logistic regression with iterative resampling to investigate the relationship between these variables occurring >1 year ago (ie, before the pandemic) and the likelihood of contracting COVID-19. Results: We observed a significantly higher frequency of self-reported violent ideation, disruptive behavior, and psychotic symptoms, for all 3 time windows of people who tested positive or were diagnosed with COVID-19 by a clinician. Using multivariable logistic regression, we observed 72% to 94% model accuracy for an increased incidence of COVID-19 in participants who reported violent ideation, disruptive behavior, or psychotic symptoms >1 year ago. Conclusions: This preliminary study found that people who reported a test or clinician diagnosis of COVID-19 also reported higher frequencies of violent ideation, disruptive behavior, or psychotic symptoms across multiple time windows, indicating that they were not likely to be the result of COVID-19. In parallel, participants who reported these behaviors >1 year ago (ie, before the pandemic) were more likely to be diagnosed with COVID-19, suggesting that violent ideation, disruptive behavior, in addition to psychotic symptoms, were associated with COVID-19 with an approximately 70% to 90% likelihood.

@article{Yunan2022,
abstract = {The relationship of human brain structure to cognitive function is complex, and how this relationship differs between childhood and adulthood is poorly understood. One strong hypothesis suggests the cognitive function of Fluid Intelligence (Gf) is dependent on prefrontal cortex and parietal cortex. In this work, we developed a novel graph convolutional neural networks (gCNNs) for the analysis of localized anatomic shape and prediction of Gf. Morphologic information of the cortical ribbons and subcortical structures was extracted from T1-weighted MRIs within two independent cohorts, the Adolescent Brain Cognitive Development Study (ABCD; age: 9.93 ± 0.62 years) of children and the Human Connectome Project (HCP; age: 28.81 ± 3.70 years). Prediction combining cortical and subcortical surfaces together yielded the highest accuracy of Gf for both ABCD (R = 0.314) and HCP datasets (R = 0.454), outperforming the state-of-the-art prediction of Gf from any other brain measures in the literature. Across both datasets, the morphology of the amygdala, hippocampus, and nucleus accumbens, along with temporal, parietal and cingulate cortex consistently drove the prediction of Gf, suggesting a significant reframing of the relationship between brain morphology and Gf to include systems involved with reward/aversion processing, judgment and decision-making, motivation, and emotion.},
author = {Wu, Yunan and Besson, Pierre and Azcona, Emanuel A. and Bandt, S. Kathleen and Parrish, Todd B. and Breiter, Hans C. and Katsaggelos, Aggelos K.},
doi = {10.1038/s41598-022-22313-x},
issn = {2045-2322},
journal = {Scientific Reports},
month = {oct},
number = {1},
pages = {17760},
pmid = {36273036},
title = {{A multicohort geometric deep learning study of age dependent cortical and subcortical morphologic interactions for fluid intelligence prediction}},
url = {https://www.nature.com/articles/s41598-022-22313-x},
volume = {12},
year = {2022}
}

The relationship of human brain structure to cognitive function is complex, and how this relationship differs between childhood and adulthood is poorly understood. One strong hypothesis suggests the cognitive function of Fluid Intelligence (Gf) is dependent on prefrontal cortex and parietal cortex. In this work, we developed a novel graph convolutional neural networks (gCNNs) for the analysis of localized anatomic shape and prediction of Gf. Morphologic information of the cortical ribbons and subcortical structures was extracted from T1-weighted MRIs within two independent cohorts, the Adolescent Brain Cognitive Development Study (ABCD; age: 9.93 ± 0.62 years) of children and the Human Connectome Project (HCP; age: 28.81 ± 3.70 years). Prediction combining cortical and subcortical surfaces together yielded the highest accuracy of Gf for both ABCD (R = 0.314) and HCP datasets (R = 0.454), outperforming the state-of-the-art prediction of Gf from any other brain measures in the literature. Across both datasets, the morphology of the amygdala, hippocampus, and nucleus accumbens, along with temporal, parietal and cingulate cortex consistently drove the prediction of Gf, suggesting a significant reframing of the relationship between brain morphology and Gf to include systems involved with reward/aversion processing, judgment and decision-making, motivation, and emotion.

@article{Miguel2022,
abstract = {Background and objective: Intracranial hemorrhage (ICH) is a life-threatening emergency that can lead to brain damage or death, with high rates of mortality and morbidity. The fast and accurate detection of ICH is important for the patient to get an early and efficient treatment. To improve this diagnostic process, the application of Deep Learning (DL) models on head CT scans is an active area of research. Although promising results have been obtained, many of the proposed models require slice-level annotations by radiologists, which are costly and time-consuming. Methods: We formulate the ICH detection as a problem of Multiple Instance Learning (MIL) that allows training with only scan-level annotations. We develop a new probabilistic method based on Deep Gaussian Processes (DGP) that is able to train with this MIL setting and accurately predict ICH at both slice- and scan-level. The proposed DGPMIL model is able to capture complex feature relations by using multiple Gaussian Process (GP) layers, as we show experimentally. Results: To highlight the advantages of DGPMIL in a general MIL setting, we first conduct several controlled experiments on the MNIST dataset. We show that multiple GP layers outperform one-layer GP models, especially for complex feature distributions. For ICH detection experiments, we use two public brain CT datasets (RSNA and CQ500). We first train a Convolutional Neural Network (CNN) with an attention mechanism to extract the image features, which are fed into our DGPMIL model to perform the final predictions. The results show that DGPMIL model outperforms VGPMIL as well as the attention-based CNN for MIL and other state-of-the-art methods for this problem. The best performing DGPMIL model reaches an AUC-ROC of 0.957 (resp. 0.909) and an AUC-PR of 0.961 (resp. 0.889) on the RSNA (resp. CQ500) dataset. Conclusion: The competitive performance at slice- and scan-level shows that DGPMIL model provides an accurate diagnosis on slices without the need for slice-level annotations by radiologists during training. As MIL is a common problem setting, our model can be applied to a broader range of other tasks, especially in medical image classification, where it can help the diagnostic process.},
author = {L{\'{o}}pez-P{\'{e}}rez, Miguel and Schmidt, Arne and Wu, Yunan and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.cmpb.2022.106783},
issn = {01692607},
journal = {Computer Methods and Programs in Biomedicine},
keywords = {Deep Gaussian processes,Intracranial hemorrhage detection,Multiple instance learning,Weakly supervised learning},
month = {jun},
pages = {106783},
pmid = {35390723},
title = {{Deep Gaussian processes for multiple instance learning: Application to CT intracranial hemorrhage detection}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0169260722001699},
volume = {219},
year = {2022}
}

Background and objective: Intracranial hemorrhage (ICH) is a life-threatening emergency that can lead to brain damage or death, with high rates of mortality and morbidity. The fast and accurate detection of ICH is important for the patient to get an early and efficient treatment. To improve this diagnostic process, the application of Deep Learning (DL) models on head CT scans is an active area of research. Although promising results have been obtained, many of the proposed models require slice-level annotations by radiologists, which are costly and time-consuming. Methods: We formulate the ICH detection as a problem of Multiple Instance Learning (MIL) that allows training with only scan-level annotations. We develop a new probabilistic method based on Deep Gaussian Processes (DGP) that is able to train with this MIL setting and accurately predict ICH at both slice- and scan-level. The proposed DGPMIL model is able to capture complex feature relations by using multiple Gaussian Process (GP) layers, as we show experimentally. Results: To highlight the advantages of DGPMIL in a general MIL setting, we first conduct several controlled experiments on the MNIST dataset. We show that multiple GP layers outperform one-layer GP models, especially for complex feature distributions. For ICH detection experiments, we use two public brain CT datasets (RSNA and CQ500). We first train a Convolutional Neural Network (CNN) with an attention mechanism to extract the image features, which are fed into our DGPMIL model to perform the final predictions. The results show that DGPMIL model outperforms VGPMIL as well as the attention-based CNN for MIL and other state-of-the-art methods for this problem. The best performing DGPMIL model reaches an AUC-ROC of 0.957 (resp. 0.909) and an AUC-PR of 0.961 (resp. 0.889) on the RSNA (resp. CQ500) dataset. Conclusion: The competitive performance at slice- and scan-level shows that DGPMIL model provides an accurate diagnosis on slices without the need for slice-level annotations by radiologists during training. As MIL is a common problem setting, our model can be applied to a broader range of other tasks, especially in medical image classification, where it can help the diagnostic process.

@article{Woodward2022,
abstract = {Background: Although the mental health impacts of COVID-19 on the general population have been well studied, studies of the long-term impacts of COVID-19 on infected individuals are relatively new. To date, depression, anxiety, and neurological symptoms associated with post-COVID-19 syndrome (PCS) have been observed in the months following COVID-19 recovery. Suicidal thoughts and behavior (STB) have also been preliminarily proposed as sequelae of COVID-19. Objective: We asked 3 questions. First, do participants reporting a history of COVID-19 diagnosis or a close relative having severe COVID-19 symptoms score higher on depression (Patient Health Questionnaire-9 [PHQ-9]) or state anxiety (State Trait Anxiety Index) screens than those who do not? Second, do participants reporting a COVID-19 diagnosis score higher on PCS-related PHQ-9 items? Third, do participants reporting a COVID-19 diagnosis or a close relative having severe COVID-19 symptoms score higher in STB before, during, or after the first year of the pandemic? Methods: This preliminary study analyzed responses to a COVID-19 and mental health questionnaire obtained from a US population sample, whose data were collected between February 2021 and March 2021. We used the Mann-Whitney U test to detect differences in the medians of the total PHQ-9 scores, PHQ-9 component scores, and several STB scores between participants claiming a past clinician diagnosis of COVID-19 and those denying one, as well as between participants claiming severe COVID-19 symptoms in a close relative and those denying them. Where significant differences existed, we created linear regression models to predict the scores based on COVID-19 response as well as demographics to identify potential confounding factors in the Mann-Whitney relationships. Moreover, for STB scores, which corresponded to 5 questions asking about 3 different time intervals (i.e., past 1 year or more, past 1 month to 1 year, and past 1 month), we developed repeated-measures ANOVAs to determine whether scores tended to vary over time. Results: We found greater total depression (PHQ-9) and state anxiety (State Trait Anxiety Index) scores in those with COVID-19 history than those without (Bonferroni P = .001 and Bonferroni P = .004) despite a similar history of diagnosed depression and anxiety. Greater scores were noted for a subset of depression symptoms (PHQ-9 items) that overlapped with the symptoms of PCS (all Bonferroni Ps < .05). Moreover, we found greater overall STB scores in those with COVID-19 history, equally in time windows preceding, during, and proceeding infection (all Bonferroni Ps < .05). Conclusions: We confirm previous studies linking depression and anxiety diagnoses to COVID-19 recovery. Moreover, our findings suggest that depression diagnoses associated with COVID-19 history relate to PCS symptoms, and that STB associated with COVID-19 in some cases precede infection.},
author = {Woodward, Sean F. and Bari, Sumra and Vike, Nicole and Lalvani, Shamal and Stetsiv, Khrystyna and Kim, Byoung Woo and Stefanopoulos, Leandros and Maglaveras, Nicos and Breiter, Hans and Katsaggelos, Aggelos K.},
doi = {10.2196/36656},
issn = {2561-326X},
journal = {JMIR Formative Research},
keywords = {COVID-19,PHQ-9,Patient Health Questionnaire-9,STAI,State Trait Anxiety Index,depression,post-COVID-19 syndrome,suicidality},
month = {oct},
number = {10},
pages = {e36656},
publisher = {JMIR Publications Inc., Toronto, Canada},
title = {{Anxiety, Post–COVID-19 Syndrome-Related Depression, and Suicidal Thoughts and Behaviors in COVID-19 Survivors: Cross-sectional Study}},
url = {https://formative.jmir.org/2022/10/e36656},
volume = {6},
year = {2022}
}

Background: Although the mental health impacts of COVID-19 on the general population have been well studied, studies of the long-term impacts of COVID-19 on infected individuals are relatively new. To date, depression, anxiety, and neurological symptoms associated with post-COVID-19 syndrome (PCS) have been observed in the months following COVID-19 recovery. Suicidal thoughts and behavior (STB) have also been preliminarily proposed as sequelae of COVID-19. Objective: We asked 3 questions. First, do participants reporting a history of COVID-19 diagnosis or a close relative having severe COVID-19 symptoms score higher on depression (Patient Health Questionnaire-9 [PHQ-9]) or state anxiety (State Trait Anxiety Index) screens than those who do not? Second, do participants reporting a COVID-19 diagnosis score higher on PCS-related PHQ-9 items? Third, do participants reporting a COVID-19 diagnosis or a close relative having severe COVID-19 symptoms score higher in STB before, during, or after the first year of the pandemic? Methods: This preliminary study analyzed responses to a COVID-19 and mental health questionnaire obtained from a US population sample, whose data were collected between February 2021 and March 2021. We used the Mann-Whitney U test to detect differences in the medians of the total PHQ-9 scores, PHQ-9 component scores, and several STB scores between participants claiming a past clinician diagnosis of COVID-19 and those denying one, as well as between participants claiming severe COVID-19 symptoms in a close relative and those denying them. Where significant differences existed, we created linear regression models to predict the scores based on COVID-19 response as well as demographics to identify potential confounding factors in the Mann-Whitney relationships. Moreover, for STB scores, which corresponded to 5 questions asking about 3 different time intervals (i.e., past 1 year or more, past 1 month to 1 year, and past 1 month), we developed repeated-measures ANOVAs to determine whether scores tended to vary over time. Results: We found greater total depression (PHQ-9) and state anxiety (State Trait Anxiety Index) scores in those with COVID-19 history than those without (Bonferroni P = .001 and Bonferroni P = .004) despite a similar history of diagnosed depression and anxiety. Greater scores were noted for a subset of depression symptoms (PHQ-9 items) that overlapped with the symptoms of PCS (all Bonferroni Ps < .05). Moreover, we found greater overall STB scores in those with COVID-19 history, equally in time windows preceding, during, and proceeding infection (all Bonferroni Ps < .05). Conclusions: We confirm previous studies linking depression and anxiety diagnoses to COVID-19 recovery. Moreover, our findings suggest that depression diagnoses associated with COVID-19 history relate to PCS symptoms, and that STB associated with COVID-19 in some cases precede infection.

@article{Pablo2020,
abstract = {In the last years, crowdsourcing is transforming the way classification training sets are obtained. Instead of relying on a single expert annotator, crowdsourcing shares the labelling effort among a large number of collaborators. For instance, this is being applied in the laureate laser interferometer gravitational waves observatory (LIGO), in order to detect glitches which might hinder the identification of true gravitational-waves. The crowdsourcing scenario poses new challenging difficulties, as it has to deal with different opinions from a heterogeneous group of annotators with unknown degrees of expertise. Probabilistic methods, such as Gaussian processes (GP), have proven successful in modeling this setting. However, GPs do not scale up well to large data sets, which hampers their broad adoption in real-world problems (in particular LIGO). This has led to the very recent introduction of deep learning based crowdsourcing methods, which have become the state-of-the-art for this type of problems. However, the accurate uncertainty quantification provided by GPs has been partially sacrificed. This is an important aspect for astrophysicists in LIGO, since a glitch detection system should provide very accurate probability distributions of its predictions. In this work, we first leverage a standard sparse GP approximation (SVGP) to develop a GP-based crowdsourcing method that factorizes into mini-batches. This makes it able to cope with previously-prohibitive data sets. This first approach, which we refer to as scalable variational Gaussian processes for crowdsourcing (SVGPCR), brings back GP-based methods to a state-of-the-art level, and excels at uncertainty quantification. SVGPCR is shown to outperform deep learning based methods and previous probabilistic ones when applied to the LIGO data. Its behavior and main properties are carefully analyzed in a controlled experiment based on the MNIST data set. Moreover, recent GP inference techniques are also adapted to crowdsourcing and evaluated experimentally.},
archivePrefix = {arXiv},
arxivId = {1911.01915},
author = {Morales-Alvarez, Pablo and Ruiz, Pablo and Coughlin, Scott and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TPAMI.2020.3025390},
eprint = {1911.01915},
issn = {0162-8828},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
keywords = {Crowdsourcing,citizen science,deep learning,laser interferometer gravitational waves observato,scalability,sparse Gaussian processes,uncertainty quantification},
month = {mar},
number = {3},
pages = {1534--1551},
pmid = {32956038},
title = {{Scalable Variational Gaussian Processes for Crowdsourcing: Glitch Detection in LIGO}},
url = {https://ieeexplore.ieee.org/document/9201029/},
volume = {44},
year = {2022}
}

In the last years, crowdsourcing is transforming the way classification training sets are obtained. Instead of relying on a single expert annotator, crowdsourcing shares the labelling effort among a large number of collaborators. For instance, this is being applied in the laureate laser interferometer gravitational waves observatory (LIGO), in order to detect glitches which might hinder the identification of true gravitational-waves. The crowdsourcing scenario poses new challenging difficulties, as it has to deal with different opinions from a heterogeneous group of annotators with unknown degrees of expertise. Probabilistic methods, such as Gaussian processes (GP), have proven successful in modeling this setting. However, GPs do not scale up well to large data sets, which hampers their broad adoption in real-world problems (in particular LIGO). This has led to the very recent introduction of deep learning based crowdsourcing methods, which have become the state-of-the-art for this type of problems. However, the accurate uncertainty quantification provided by GPs has been partially sacrificed. This is an important aspect for astrophysicists in LIGO, since a glitch detection system should provide very accurate probability distributions of its predictions. In this work, we first leverage a standard sparse GP approximation (SVGP) to develop a GP-based crowdsourcing method that factorizes into mini-batches. This makes it able to cope with previously-prohibitive data sets. This first approach, which we refer to as scalable variational Gaussian processes for crowdsourcing (SVGPCR), brings back GP-based methods to a state-of-the-art level, and excels at uncertainty quantification. SVGPCR is shown to outperform deep learning based methods and previous probabilistic ones when applied to the LIGO data. Its behavior and main properties are carefully analyzed in a controlled experiment based on the MNIST data set. Moreover, recent GP inference techniques are also adapted to crowdsourcing and evaluated experimentally.

@article{Fernando2022,
abstract = {Stain variation between images is a main issue in the analysis of histological images. These color variations, produced by different staining protocols and scanners in each laboratory, hamper the performance of computer-aided diagnosis (CAD) systems that are usually unable to generalize to unseen color distributions. Blind color deconvolution techniques separate multi-stained images into single stained bands that can then be used to reduce the generalization error of CAD systems through stain color normalization and/or stain color augmentation. In this work, we present a Bayesian modeling and inference blind color deconvolution framework based on the K-Singular Value Decomposition algorithm. Two possible inference procedures, variational and empirical Bayes are presented. Both provide the automatic estimation of the stain color matrix, stain concentrations and all model parameters. The proposed framework is tested on stain separation, image normalization, stain color augmentation, and classification problems.},
author = {P{\'{e}}rez-Bueno, Fernando and Serra, Juan G. and Vega, Miguel and Mateos, Javier and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.compmedimag.2022.102048},
issn = {08956111},
journal = {Computerized Medical Imaging and Graphics},
keywords = {Bayesian modeling,Blind Color Deconvolution,Histological images,Stain Normalization},
month = {apr},
pages = {102048},
pmid = {35202893},
title = {{Bayesian K-SVD for H and E blind color deconvolution. Applications to stain normalization, data augmentation and cancer classification}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0895611122000210},
volume = {97},
year = {2022}
}

Stain variation between images is a main issue in the analysis of histological images. These color variations, produced by different staining protocols and scanners in each laboratory, hamper the performance of computer-aided diagnosis (CAD) systems that are usually unable to generalize to unseen color distributions. Blind color deconvolution techniques separate multi-stained images into single stained bands that can then be used to reduce the generalization error of CAD systems through stain color normalization and/or stain color augmentation. In this work, we present a Bayesian modeling and inference blind color deconvolution framework based on the K-Singular Value Decomposition algorithm. Two possible inference procedures, variational and empirical Bayes are presented. Both provide the automatic estimation of the stain color matrix, stain concentrations and all model parameters. The proposed framework is tested on stain separation, image normalization, stain color augmentation, and classification problems.

@article{ballester2022application,
abstract = {The Tauc–Lorentz–Urbach (TLU) dispersion model allows us to build a dielectric function from only a few parameters. However, this dielectric function is non-analytic and presents some mathematical drawbacks. As a consequence of this issue, the model becomes inaccurate. In the present work, we will adopt a procedure to conveniently transform the TLU model into a self-consistent dispersion model. The transformation involves the integration of the original TLU imaginary dielectric function ϵ2 by using a Lorentzian-type function of semi-width, $\Gamma$. This novel model is analytic and obeys the other necessary mathematical requirements of the optical constants of solid-state materials. The main difference with the non-analytic TLU model occurs at values of the photon energy near or lower than that of the bandgap energy (within the Urbach absorption region). In particular, this new model allows us to reliably extend the optical characterization of amorphous-semiconductor thin films within the limit to zero photon energy. To the best of our knowledge, this is the first time that the analytic TLU model has been successfully used to accurately determine the optical constants of unhydrogenated a-Si films using only their normal-incidence transmission spectra.},
author = {Ballester, Manuel and Garc{\'{i}}a, Marcos and M{\'{a}}rquez, Almudena P. and Blanco, Eduardo and Fern{\'{a}}ndez, Susana M. and Minkov, Dorian and Katsaggelos, Aggelos K. and Cossairt, Oliver and Willomitzer, Florian and M{\'{a}}rquez, Emilio},
doi = {10.3390/coatings12101549},
issn = {2079-6412},
journal = {Coatings},
keywords = {Tauc–Lorentz model,Tauc–Lorentz–Urbach model,amorphous semiconductors,dielectric function,optical properties,thin-film characterization},
month = {oct},
number = {10},
pages = {1549},
publisher = {MDPI},
title = {{Application of the Holomorphic Tauc-Lorentz-Urbach Function to Extract the Optical Constants of Amorphous Semiconductor Thin Films}},
url = {https://www.mdpi.com/2079-6412/12/10/1549},
volume = {12},
year = {2022}
}

The Tauc–Lorentz–Urbach (TLU) dispersion model allows us to build a dielectric function from only a few parameters. However, this dielectric function is non-analytic and presents some mathematical drawbacks. As a consequence of this issue, the model becomes inaccurate. In the present work, we will adopt a procedure to conveniently transform the TLU model into a self-consistent dispersion model. The transformation involves the integration of the original TLU imaginary dielectric function ϵ2 by using a Lorentzian-type function of semi-width, $Γ$. This novel model is analytic and obeys the other necessary mathematical requirements of the optical constants of solid-state materials. The main difference with the non-analytic TLU model occurs at values of the photon energy near or lower than that of the bandgap energy (within the Urbach absorption region). In particular, this new model allows us to reliably extend the optical characterization of amorphous-semiconductor thin films within the limit to zero photon energy. To the best of our knowledge, this is the first time that the analytic TLU model has been successfully used to accurately determine the optical constants of unhydrogenated a-Si films using only their normal-incidence transmission spectra.

@article{Jennifer2022,
abstract = {Quantifying motion in 3D is important for studying the behavior of humans and other animals, but manual pose annotations are expensive and time-consuming to obtain. Self-supervised keypoint discovery is a promising strategy for estimating 3D poses without annotations. However, current keypoint discovery approaches commonly process single 2D views and do not operate in the 3D space. We propose a new method to perform self-supervised keypoint discovery in 3D from multi-view videos of behaving agents, without any keypoint or bounding box supervision in 2D or 3D. Our method uses an encoder-decoder architecture with a 3D volumetric heatmap, trained to reconstruct spatiotemporal differences across multiple views, in addition to joint length constraints on a learned 3D skeleton of the subject. In this way, we discover keypoints without requiring manual supervision in videos of humans and rats, demonstrating the potential of 3D keypoint discovery for studying behavior.},
archivePrefix = {arXiv},
arxivId = {2212.07401},
author = {Sun, Jennifer J. and Karashchuk, Pierre and Dravid, Amil and Ryou, Serim and Fereidooni, Sonia and Tuthill, John and Katsaggelos, Aggelos and Brunton, Bingni W. and Gkioxari, Georgia and Kennedy, Ann and Yue, Yisong and Perona, Pietro},
eprint = {2212.07401},
journal = {arXiv preprint arXiv:2212.07401},
month = {dec},
title = {{BKinD-3D: Self-Supervised 3D Keypoint Discovery from Multi-View Videos}},
url = {http://arxiv.org/abs/2212.07401},
year = {2022}
}

Quantifying motion in 3D is important for studying the behavior of humans and other animals, but manual pose annotations are expensive and time-consuming to obtain. Self-supervised keypoint discovery is a promising strategy for estimating 3D poses without annotations. However, current keypoint discovery approaches commonly process single 2D views and do not operate in the 3D space. We propose a new method to perform self-supervised keypoint discovery in 3D from multi-view videos of behaving agents, without any keypoint or bounding box supervision in 2D or 3D. Our method uses an encoder-decoder architecture with a 3D volumetric heatmap, trained to reconstruct spatiotemporal differences across multiple views, in addition to joint length constraints on a learned 3D skeleton of the subject. In this way, we discover keypoints without requiring manual supervision in videos of humans and rats, demonstrating the potential of 3D keypoint discovery for studying behavior.

@article{chopp2022image,
abstract = {X-ray fluorescence (XRF) analysis of art objects has rapidly gained popularity since the late 2000s due to its increased accessibility to scientists. This introduced an imaging component whereby the XRF image volume provides clues as to which chemical elements are present and where they are located spatially in the object. However, as is the nature of collecting measurements, there are limitations preventing perfect acquisition; e.g, spatial resolution, signal-to-noise ratio, etc. The field of image processing, in part, aims to overcome these limitations. Image processing applications in XRF imaging are only just starting to arise due to the increased interest and availability in XRF analysis. In this article, we aim to reach readers in XRF imaging or image processing in an effort to call for further research in the field. We review the basics of XRF imaging and analysis that is tailored for those unfamiliar with this imaging modality. We then delve into various publications of image processing methods as applied to XRF data. Throughout this article, we examine (and opine on) the XRF field through a lens of the image processing field.},
author = {Chopp, Henry and McGeachy, Alicia and Alfeld, Matthias and Cossairt, Oliver and Walton, Marc and Katsaggelos, Aggelos},
doi = {10.1109/MBITS.2022.3197100},
issn = {2692-4110},
journal = {IEEE BITS the Information Theory Magazine},
number = {1},
pages = {1--12},
publisher = {IEEE},
title = {{Image Processing Perspectives of X-Ray Fluorescence Data in Cultural Heritage Sciences}},
url = {https://ieeexplore.ieee.org/document/9851850/},
volume = {2},
year = {2022}
}

X-ray fluorescence (XRF) analysis of art objects has rapidly gained popularity since the late 2000s due to its increased accessibility to scientists. This introduced an imaging component whereby the XRF image volume provides clues as to which chemical elements are present and where they are located spatially in the object. However, as is the nature of collecting measurements, there are limitations preventing perfect acquisition; e.g, spatial resolution, signal-to-noise ratio, etc. The field of image processing, in part, aims to overcome these limitations. Image processing applications in XRF imaging are only just starting to arise due to the increased interest and availability in XRF analysis. In this article, we aim to reach readers in XRF imaging or image processing in an effort to call for further research in the field. We review the basics of XRF imaging and analysis that is tailored for those unfamiliar with this imaging modality. We then delve into various publications of image processing methods as applied to XRF data. Throughout this article, we examine (and opine on) the XRF field through a lens of the image processing field.

@article{liu2022digital,
abstract = {Digital restoration is a rapidly growing methodology within the field of heritage conservation, especially for early cinematic films which have intrinsically unstable dye colourants that suffer from irreversible colour fading. Although numerous techniques to restore film digitally have emerged recently, complex degradation remains a challenging problem. This paper proposes a novel vector quantization (VQ) algorithm for restoring movie frames based on the acquisition of spectroscopic data with a custom-made push-broom VNIR hyperspectral camera (380–780 nm). The VQ algorithm utilizes what we call a multi-codebook that correlates degraded areas with corresponding non-degraded ones selected from reference frames. The spectral-codebook was compared with a professional commercially available film restoration software (DaVinci Resolve 17) tested both on RGB and on hyperspectral providing better results in terms of colour reconstruction.},
author = {Liu, L. and Catelli, E. and Katsaggelos, A. and Sciutto, G. and Mazzeo, R. and Milanic, M. and Stergar, J. and Prati, S. and Walton, M.},
doi = {10.1038/s41598-022-25248-5},
issn = {2045-2322},
journal = {Scientific Reports},
month = {dec},
number = {1},
pages = {21982},
pmid = {36539479},
publisher = {Nature Publishing Group UK London},
title = {{Digital restoration of colour cinematic films using imaging spectroscopy and machine learning}},
url = {https://www.nature.com/articles/s41598-022-25248-5},
volume = {12},
year = {2022}
}

Digital restoration is a rapidly growing methodology within the field of heritage conservation, especially for early cinematic films which have intrinsically unstable dye colourants that suffer from irreversible colour fading. Although numerous techniques to restore film digitally have emerged recently, complex degradation remains a challenging problem. This paper proposes a novel vector quantization (VQ) algorithm for restoring movie frames based on the acquisition of spectroscopic data with a custom-made push-broom VNIR hyperspectral camera (380–780 nm). The VQ algorithm utilizes what we call a multi-codebook that correlates degraded areas with corresponding non-degraded ones selected from reference frames. The spectral-codebook was compared with a professional commercially available film restoration software (DaVinci Resolve 17) tested both on RGB and on hyperspectral providing better results in terms of colour reconstruction.

@article{dai2022adaptive,
abstract = {Dense depth map capture is challenging in existing active sparse illumination based depth acquisition techniques, such as LiDAR. Various techniques have been proposed to estimate a dense depth map based on fusion of the sparse depth map measurement with the RGB image. Recent advances in hardware enable adaptive depth measurements resulting in further improvement of the dense depth map estimation. In this paper, we study the topic of estimating dense depth from depth sampling. The adaptive sparse depth sampling network is jointly trained with a fusion network of an RGB image and sparse depth, to generate optimal adaptive sampling masks. Deep learning based superpixel sampling and soft sampling approximation are applied. We show that such adaptive sampling masks can generalize well to many RGB and sparse depth fusion algorithms under a variety of sampling rates (as low as 0.0625%). The proposed adaptive sampling method is fully differentiable and flexible to be trained end-to-end with upstream perception algorithms.},
author = {Dai, Qiqin and Li, Fengqiang and Cossairt, Oliver and Katsaggelos, Aggelos K.},
doi = {10.1109/TCI.2022.3155377},
issn = {2333-9403},
journal = {IEEE Transactions on Computational Imaging},
keywords = {Adaptive sampling,Deep learning,Depth estimation,Sensor fusion},
pages = {224--235},
publisher = {IEEE},
title = {{Adaptive Illumination Based Depth Sensing Using Deep Superpixel and Soft Sampling Approximation}},
url = {https://ieeexplore.ieee.org/document/9723580/},
volume = {8},
year = {2022}
}

Dense depth map capture is challenging in existing active sparse illumination based depth acquisition techniques, such as LiDAR. Various techniques have been proposed to estimate a dense depth map based on fusion of the sparse depth map measurement with the RGB image. Recent advances in hardware enable adaptive depth measurements resulting in further improvement of the dense depth map estimation. In this paper, we study the topic of estimating dense depth from depth sampling. The adaptive sparse depth sampling network is jointly trained with a fusion network of an RGB image and sparse depth, to generate optimal adaptive sampling masks. Deep learning based superpixel sampling and soft sampling approximation are applied. We show that such adaptive sampling masks can generalize well to many RGB and sparse depth fusion algorithms under a variety of sampling rates (as low as 0.0625%). The proposed adaptive sampling method is fully differentiable and flexible to be trained end-to-end with upstream perception algorithms.

@article{Vassilis2022,
abstract = {Monitoring and treatment of severely ill COVID-19 patients in the ICU poses many challenges. The effort to understand the pathophysiology and progress of the disease requires high-quality annotated multi-parameter databases. We present CoCross, a platform that enables the monitoring and fusion of clinical information from in-ICU COVID-19 patients into an annotated database. CoCross consists of three components: (1) The CoCross4Pros native android application, a modular application, managing the interaction with portable medical devices, (2) the cloud-based data management services built-upon HL7 FHIR and ontologies, (3) the web-based application for intensivists, providing real-time review and analytics of the acquired measurements and auscultations. The platform has been successfully deployed since June 2020 in two ICUs in Greece resulting in a dynamic unified annotated database integrating clinical information with chest sounds and diagnostic imaging. Until today multisource data from 176 ICU patients were acquired and imported in the CoCross database, corresponding to a five-day average monitoring period including a dataset with 3477 distinct auscultations. The platform is well accepted and positively rated by the users regarding the overall experience.},
author = {Kilintzis, Vassilis and Beredimas, Nikolaos and Kaimakamis, Evangelos and Stefanopoulos, Leandros and Chatzis, Evangelos and Jahaj, Edison and Bitzani, Militsa and Kotanidou, Anastasia and Katsaggelos, Aggelos K. and Maglaveras, Nicos},
doi = {10.3390/healthcare10020276},
issn = {2227-9032},
journal = {Healthcare},
keywords = {Biomedical monitoring,COVID-19,Heart sounds,ICT pl,ICT platform,ICU,Lung sounds,Lung ultrasound,Research database,X-rays},
month = {jan},
number = {2},
pages = {276},
title = {{CoCross: An ICT Platform Enabling Monitoring Recording and Fusion of Clinical Information Chest Sounds and Imaging of COVID-19 ICU Patients}},
url = {https://www.mdpi.com/2227-9032/10/2/276},
volume = {10},
year = {2022}
}

Monitoring and treatment of severely ill COVID-19 patients in the ICU poses many challenges. The effort to understand the pathophysiology and progress of the disease requires high-quality annotated multi-parameter databases. We present CoCross, a platform that enables the monitoring and fusion of clinical information from in-ICU COVID-19 patients into an annotated database. CoCross consists of three components: (1) The CoCross4Pros native android application, a modular application, managing the interaction with portable medical devices, (2) the cloud-based data management services built-upon HL7 FHIR and ontologies, (3) the web-based application for intensivists, providing real-time review and analytics of the acquired measurements and auscultations. The platform has been successfully deployed since June 2020 in two ICUs in Greece resulting in a dynamic unified annotated database integrating clinical information with chest sounds and diagnostic imaging. Until today multisource data from 176 ICU patients were acquired and imported in the CoCross database, corresponding to a five-day average monitoring period including a dataset with 3477 distinct auscultations. The platform is well accepted and positively rated by the users regarding the overall experience.

@article{Petmezas2022,
abstract = {Background: Electrocardiogram (ECG) is one of the most common noninvasive diagnostic tools that can provide useful information regarding a patient's health status. Deep learning (DL) is an area of intense exploration that leads the way in most attempts to create powerful diagnostic models based on physiological signals. Objective: This study aimed to provide a systematic review of DL methods applied to ECG data for various clinical applications. Methods: The PubMed search engine was systematically searched by combining “deep learning” and keywords such as “ecg,” “ekg,” “electrocardiogram,” “electrocardiography,” and “electrocardiology.” Irrelevant articles were excluded from the study after screening titles and abstracts, and the remaining articles were further reviewed. The reasons for article exclusion were manuscripts written in any language other than English, absence of ECG data or DL methods involved in the study, and absence of a quantitative evaluation of the proposed approaches. Results: We identified 230 relevant articles published between January 2020 and December 2021 and grouped them into 6 distinct medical applications, namely, blood pressure estimation, cardiovascular disease diagnosis, ECG analysis, biometric recognition, sleep analysis, and other clinical analyses. We provide a complete account of the state-of-the-art DL strategies per the field of application, as well as major ECG data sources. We also present open research problems, such as the lack of attempts to address the issue of blood pressure variability in training data sets, and point out potential gaps in the design and implementation of DL models. Conclusions: We expect that this review will provide insights into state-of-the-art DL methods applied to ECG data and point to future directions for research on DL to create robust models that can assist medical experts in clinical decision-making.},
author = {Petmezas, Georgios and Stefanopoulos, Leandros and Kilintzis, Vassilis and Tzavelis, Andreas and Rogers, John A. and Katsaggelos, Aggelos K. and Maglaveras, Nicos},
doi = {10.2196/38454},
issn = {2291-9694},
journal = {JMIR Medical Informatics},
keywords = {CNN,ECG,ECG databases,LSTM,ResNet,clinical decision,convolutional neural networks,decision support,deep learning,diagnostic tools,electrocardiogram,long short-term memory,residual neural network},
month = {aug},
number = {8},
pages = {e38454},
publisher = {JMIR Publications Toronto, Canada},
title = {{State-of-the-Art Deep Learning Methods on Electrocardiogram Data: Systematic Review}},
url = {https://medinform.jmir.org/2022/8/e38454},
volume = {10},
year = {2022}
}

Background: Electrocardiogram (ECG) is one of the most common noninvasive diagnostic tools that can provide useful information regarding a patient's health status. Deep learning (DL) is an area of intense exploration that leads the way in most attempts to create powerful diagnostic models based on physiological signals. Objective: This study aimed to provide a systematic review of DL methods applied to ECG data for various clinical applications. Methods: The PubMed search engine was systematically searched by combining “deep learning” and keywords such as “ecg,” “ekg,” “electrocardiogram,” “electrocardiography,” and “electrocardiology.” Irrelevant articles were excluded from the study after screening titles and abstracts, and the remaining articles were further reviewed. The reasons for article exclusion were manuscripts written in any language other than English, absence of ECG data or DL methods involved in the study, and absence of a quantitative evaluation of the proposed approaches. Results: We identified 230 relevant articles published between January 2020 and December 2021 and grouped them into 6 distinct medical applications, namely, blood pressure estimation, cardiovascular disease diagnosis, ECG analysis, biometric recognition, sleep analysis, and other clinical analyses. We provide a complete account of the state-of-the-art DL strategies per the field of application, as well as major ECG data sources. We also present open research problems, such as the lack of attempts to address the issue of blood pressure variability in training data sets, and point out potential gaps in the design and implementation of DL models. Conclusions: We expect that this review will provide insights into state-of-the-art DL methods applied to ECG data and point to future directions for research on DL to create robust models that can assist medical experts in clinical decision-making.

@article{Michael2022a,
abstract = {BACKGROUND AND PURPOSE: Prioritizing reading of noncontrast head CT examinations through an automated triage system may improve time to care for patients with acute neuroradiologic findings. We present a natural language-processing approach for labeling findings in noncontrast head CT reports, which permits creation of a large, labeled dataset of head CT images for development of emergent-finding detection and reading-prioritization algorithms. MATERIALS AND METHODS: In this retrospective study, 1002 clinical radiology reports from noncontrast head CTs collected between 2008 and 2013 were manually labeled across 12 common neuroradiologic finding categories. Each report was then encoded using an n-gram model of unigrams, bigrams, and trigrams. A logistic regression model was then trained to label each report for every common finding. Models were trained and assessed using a combination of L2 regularization and 5-fold cross-validation. RESULTS: Model performance was strongest for the fracture, hemorrhage, herniation, mass effect, pneumocephalus, postoperative status, and volume loss models in which the area under the receiver operating characteristic curve exceeded 0.95. Performance was relatively weaker for the edema, hydrocephalus, infarct, tumor, and white-matter disease models (area under the receiver operating characteristic curve > 0.85). Analysis of coefficients revealed finding-specific words among the top coefficients in each model. Class output probabilities were found to be a useful indicator of predictive error on individual report examples in higher-performing models. CONCLUSIONS: Combining logistic regression with n-gram encoding is a robust approach to labeling common findings in noncontrast head CT reports.},
author = {Iorga, Michael and Drakopoulos, M. and Naidech, A.M. and Katsaggelos, A.K. and Parrish, T.B. and Hill, V.B.},
doi = {10.3174/ajnr.A7500},
issn = {0195-6108},
journal = {American Journal of Neuroradiology},
month = {may},
number = {5},
pages = {721--726},
pmid = {35483905},
title = {{Labeling Noncontrast Head CT Reports for Common Findings Using Natural Language Processing}},
url = {http://www.ajnr.org/lookup/doi/10.3174/ajnr.A7500},
volume = {43},
year = {2022}
}

BACKGROUND AND PURPOSE: Prioritizing reading of noncontrast head CT examinations through an automated triage system may improve time to care for patients with acute neuroradiologic findings. We present a natural language-processing approach for labeling findings in noncontrast head CT reports, which permits creation of a large, labeled dataset of head CT images for development of emergent-finding detection and reading-prioritization algorithms. MATERIALS AND METHODS: In this retrospective study, 1002 clinical radiology reports from noncontrast head CTs collected between 2008 and 2013 were manually labeled across 12 common neuroradiologic finding categories. Each report was then encoded using an n-gram model of unigrams, bigrams, and trigrams. A logistic regression model was then trained to label each report for every common finding. Models were trained and assessed using a combination of L2 regularization and 5-fold cross-validation. RESULTS: Model performance was strongest for the fracture, hemorrhage, herniation, mass effect, pneumocephalus, postoperative status, and volume loss models in which the area under the receiver operating characteristic curve exceeded 0.95. Performance was relatively weaker for the edema, hydrocephalus, infarct, tumor, and white-matter disease models (area under the receiver operating characteristic curve > 0.85). Analysis of coefficients revealed finding-specific words among the top coefficients in each model. Class output probabilities were found to be a useful indicator of predictive error on individual report examples in higher-performing models. CONCLUSIONS: Combining logistic regression with n-gram encoding is a robust approach to labeling common findings in noncontrast head CT reports.

@article{Petmezas2022a,
abstract = {Respiratory diseases constitute one of the leading causes of death worldwide and directly affect the patient's quality of life. Early diagnosis and patient monitoring, which conventionally include lung auscultation, are essential for the efficient management of respiratory diseases. Manual lung sound interpretation is a subjective and time-consuming process that requires high medical expertise. The capabilities that deep learning offers could be exploited in order that robust lung sound classification models can be designed. In this paper, we propose a novel hybrid neural model that implements the focal loss (FL) function to deal with training data imbalance. Features initially extracted from short-time Fourier transform (STFT) spectrograms via a convolutional neural network (CNN) are given as input to a long short-term memory (LSTM) network that memorizes the temporal dependencies between data and classifies four types of lung sounds, including normal, crackles, wheezes, and both crackles and wheezes. The model was trained and tested on the ICBHI 2017 Respiratory Sound Database and achieved state-of-the-art results using three different data splitting strategies—namely, sensitivity 47.37%, specificity 82.46%, score 64.92% and accuracy 73.69% for the official 60/40 split, sensitivity 52.78%, specificity 84.26%, score 68.52% and accuracy 76.39% using interpatient 10-fold cross validation, and sensitivity 60.29% and accuracy 74.57% using leave-one-out cross validation.},
author = {Petmezas, Georgios and Cheimariotis, Grigorios-Aris and Stefanopoulos, Leandros and Rocha, Bruno and Paiva, Rui Pedro and Katsaggelos, Aggelos K. and Maglaveras, Nicos},
doi = {10.3390/s22031232},
issn = {1424-8220},
journal = {Sensors},
keywords = {Asthma,CNN,COPD,Crackles,Focal loss,LSTM,Lung sounds,STFT,Wheezes},
month = {feb},
number = {3},
pages = {1232},
pmid = {35161977},
publisher = {MDPI},
title = {{Automated Lung Sound Classification Using a Hybrid CNN-LSTM Network and Focal Loss Function}},
url = {https://www.mdpi.com/1424-8220/22/3/1232},
volume = {22},
year = {2022}
}

Respiratory diseases constitute one of the leading causes of death worldwide and directly affect the patient's quality of life. Early diagnosis and patient monitoring, which conventionally include lung auscultation, are essential for the efficient management of respiratory diseases. Manual lung sound interpretation is a subjective and time-consuming process that requires high medical expertise. The capabilities that deep learning offers could be exploited in order that robust lung sound classification models can be designed. In this paper, we propose a novel hybrid neural model that implements the focal loss (FL) function to deal with training data imbalance. Features initially extracted from short-time Fourier transform (STFT) spectrograms via a convolutional neural network (CNN) are given as input to a long short-term memory (LSTM) network that memorizes the temporal dependencies between data and classifies four types of lung sounds, including normal, crackles, wheezes, and both crackles and wheezes. The model was trained and tested on the ICBHI 2017 Respiratory Sound Database and achieved state-of-the-art results using three different data splitting strategies—namely, sensitivity 47.37%, specificity 82.46%, score 64.92% and accuracy 73.69% for the official 60/40 split, sensitivity 52.78%, specificity 84.26%, score 68.52% and accuracy 76.39% using interpatient 10-fold cross validation, and sensitivity 60.29% and accuracy 74.57% using leave-one-out cross validation.

@article{Emanuel2022,
abstract = {Operant keypress tasks, where each action has a consequence, have been analogized to the construct of "wanting" and produce lawful relationships in humans that quantify preferences for approach and avoidance behavior. It is unknown if rating tasks without an operant framework, which can be analogized to "liking", show similar lawful relationships. We studied three independent cohorts of participants (N = 501, 506, and 4,019 participants) collected by two distinct organizations, using the same 7-point Likert scale to rate negative to positive preferences for pictures from the International Affective Picture Set. Picture ratings without an operant framework produced similar value functions, limit functions, and trade-off functions to those reported in the literature for operant keypress tasks, all with goodness of fits above 0.75. These value, limit, and trade-off functions were discrete in their mathematical formulation, recurrent across all three independent cohorts, and demonstrated scaling between individual and group curves. In all three experiments, the computation of loss aversion showed 95% confidence intervals below the value of 2, arguing against a strong overweighting of losses relative to gains, as has previously been reported for keypress tasks or games of chance with calibrated uncertainty. Graphed features from the three cohorts were similar and argue that preference assessments meet three of four criteria for lawfulness, providing a simple, short, and low-cost method for the quantitative assessment of preference without forced choice decisions, games of chance, or operant keypressing. This approach can easily be implemented on any digital device with a screen (e.g., cellphones).},
archivePrefix = {arXiv},
arxivId = {2203.06448},
author = {Azcona, Emanuel A. and Kim, Byoung-Woo and Vike, Nicole L. and Bari, Sumra and Lalvani, Shamal and Stefanopoulos, Leandros and Woodward, Sean and Block, Martin and Katsaggelos, Aggelos K. and Breiter, Hans C.},
eprint = {2203.06448},
journal = {arXiv preprint arXiv:2203.06448},
month = {mar},
title = {{Discrete, recurrent, and scalable patterns in human judgement underlie affective picture ratings}},
url = {http://arxiv.org/abs/2203.06448},
year = {2022}
}

Operant keypress tasks, where each action has a consequence, have been analogized to the construct of "wanting" and produce lawful relationships in humans that quantify preferences for approach and avoidance behavior. It is unknown if rating tasks without an operant framework, which can be analogized to "liking", show similar lawful relationships. We studied three independent cohorts of participants (N = 501, 506, and 4,019 participants) collected by two distinct organizations, using the same 7-point Likert scale to rate negative to positive preferences for pictures from the International Affective Picture Set. Picture ratings without an operant framework produced similar value functions, limit functions, and trade-off functions to those reported in the literature for operant keypress tasks, all with goodness of fits above 0.75. These value, limit, and trade-off functions were discrete in their mathematical formulation, recurrent across all three independent cohorts, and demonstrated scaling between individual and group curves. In all three experiments, the computation of loss aversion showed 95% confidence intervals below the value of 2, arguing against a strong overweighting of losses relative to gains, as has previously been reported for keypress tasks or games of chance with calibrated uncertainty. Graphed features from the three cohorts were similar and argue that preference assessments meet three of four criteria for lawfulness, providing a simple, short, and low-cost method for the quantitative assessment of preference without forced choice decisions, games of chance, or operant keypressing. This approach can easily be implemented on any digital device with a screen (e.g., cellphones).

@article{BingjieJenny2022,
abstract = {X-ray fluorescence spectroscopy (XRF) plays an important role for elemental analysis in a wide range of scientific fields, especially in cultural heritage.},
archivePrefix = {arXiv},
arxivId = {2207.12651},
author = {Xu, Bingjie Jenny and Wu, Yunan and Hao, Pengxiao and Vermeulen, Marc and McGeachy, Alicia and Smith, Kate and Eremin, Katherine and Rayner, Georgina and Verri, Giovanni and Willomitzer, Florian and Alfeld, Matthias and Tumblin, Jack and Katsaggelos, Aggelos and Walton, Marc},
doi = {10.1039/D2JA00246A},
eprint = {2207.12651},
issn = {0267-9477},
journal = {Journal of Analytical Atomic Spectrometry},
number = {12},
pages = {2672--2682},
title = {{Can deep learning assist automatic identification of layered pigments from XRF data?}},
url = {http://xlink.rsc.org/?DOI=D2JA00246A},
volume = {37},
year = {2022}
}

X-ray fluorescence spectroscopy (XRF) plays an important role for elemental analysis in a wide range of scientific fields, especially in cultural heritage.

@article{Xin2022,
author = {Yuan, Xin and Brady, David J. and Suo, Jinli and Arguello, Henry and Rodrigues, Miguel and Katsaggelos, Aggelos K.},
doi = {10.1109/JSTSP.2022.3185190},
issn = {1932-4553},
journal = {IEEE Journal of Selected Topics in Signal Processing},
month = {jun},
number = {4},
pages = {603--607},
title = {{Editorial: Introduction to the Special Issue on Deep Learning for High-Dimensional Sensing}},
url = {https://ieeexplore.ieee.org/document/9829820/},
volume = {16},
year = {2022}
}

@article{banerjee2021learning,
abstract = {Roomerature semiconductor radiation detectors (RTSDs) such as CdTe are becoming popular in computed tomography (CT) imaging. These detectors are often pixelated, requiring cumbersome postinteraction 3-D event reconstruction, which can benefit from detailed material characterization at the micron level. Transport properties and material defects with respect to electrons and holes are to be characterized, which is a labor-intensive process. Current state-of-the-art characterization is done either as a whole or at most pixel-by-pixel over the detector material. In this article, we propose a novel learning-based physical model to infer material properties at the microscopic level for RTSD. Our approach uses a novel physics-inspired learning model based on physical transport of charges with trapping centers for electrons and holes in the detector. The proposed model learns these material properties from known or measured input charges to the detector along with known or measured output signals and distributed charges in the bulk of the RTSD. The actual physical detector is divided into voxels in space and takes into account different material properties (such as drift, trapping, detrapping, and recombination) in each voxel as learnable model parameters. The model is based on a physics-inspired recurrent neural network model instead of traditional convolutional or fully connected networks. The advantage of our approach is the one-to-one relationship between the actual physical parameters of the voxels and learnable weights in the model, far fewer trainable parameters compared to traditional neural network approaches and less training time. The performance of our model has been evaluated on cadmium zinc telluride (CdZnTe), with voxels of three sizes, 25, 50, and 100$\mu$ m, for single charge input as well as multiple charge inputs at different voxel positions. Our learning-based model provides material properties with higher spatial resolution and performs well in all scenarios and matches the actual physical parameters better than state-of-the-art classical approaches.},
author = {Banerjee, Srutarshi and Rodrigues, Miesher and Vija, Alexander Hans and Katsaggelos, Aggelos K.},
doi = {10.1109/TNS.2021.3130486},
issn = {0018-9499},
journal = {IEEE Transactions on Nuclear Science},
keywords = {Charge transport,Defects,Detrapping,Learning-based model,Material characterization,Room temperature semiconductor detector,Schokley-Ramo theorem,Trapping,Trapping centers},
month = {jan},
number = {1},
pages = {2--16},
publisher = {IEEE},
title = {{A Learning-Based Physical Model of Charge Transport in Room-Temperature Semiconductor Detectors}},
url = {https://ieeexplore.ieee.org/document/9625943/},
volume = {69},
year = {2022}
}

Roomerature semiconductor radiation detectors (RTSDs) such as CdTe are becoming popular in computed tomography (CT) imaging. These detectors are often pixelated, requiring cumbersome postinteraction 3-D event reconstruction, which can benefit from detailed material characterization at the micron level. Transport properties and material defects with respect to electrons and holes are to be characterized, which is a labor-intensive process. Current state-of-the-art characterization is done either as a whole or at most pixel-by-pixel over the detector material. In this article, we propose a novel learning-based physical model to infer material properties at the microscopic level for RTSD. Our approach uses a novel physics-inspired learning model based on physical transport of charges with trapping centers for electrons and holes in the detector. The proposed model learns these material properties from known or measured input charges to the detector along with known or measured output signals and distributed charges in the bulk of the RTSD. The actual physical detector is divided into voxels in space and takes into account different material properties (such as drift, trapping, detrapping, and recombination) in each voxel as learnable model parameters. The model is based on a physics-inspired recurrent neural network model instead of traditional convolutional or fully connected networks. The advantage of our approach is the one-to-one relationship between the actual physical parameters of the voxels and learnable weights in the model, far fewer trainable parameters compared to traditional neural network approaches and less training time. The performance of our model has been evaluated on cadmium zinc telluride (CdZnTe), with voxels of three sizes, 25, 50, and 100$μ$ m, for single charge input as well as multiple charge inputs at different voxel positions. Our learning-based model provides material properties with higher spatial resolution and performs well in all scenarios and matches the actual physical parameters better than state-of-the-art classical approaches.

@article{Aggelosd,
abstract = {XRFast is a new software package written in Julia to decompose XRF imaging dataset.},
author = {Vermeulen, Marc and McGeachy, Alicia and Xu, Bingjie and Chopp, Henry and Katsaggelos, Aggelos and Meyers, Rebecca and Alfeld, Matthias and Walton, Marc},
doi = {10.1039/D2JA00114D},
issn = {0267-9477},
journal = {Journal of Analytical Atomic Spectrometry},
number = {10},
pages = {2130--2143},
title = {{XRFast a new software package for processing of MA-XRF datasets using machine learning}},
url = {https://pubs.rsc.org/en/content/articlelanding/2022/JA/D2JA00114D http://xlink.rsc.org/?DOI=D2JA00114D},
volume = {37},
year = {2022}
}

XRFast is a new software package written in Julia to decompose XRF imaging dataset.

@article{arief2022towards,
abstract = {A robust–accurate estimation of fluid flow is the main building block of a distributed virtual flow meter. Unfortunately, a big leap in algorithm development would be required for this objective to come to fruition, mainly due to the inability of current machine learning algorithms to make predictions outside the training data distribution. To improve predictions outside the training distribution, we explore the continual learning (CL) paradigm for accurately estimating the characteristics of fluid flow in pipelines. A significant challenge facing CL is the concept of catastrophic forgetting. In this paper, we provide a novel approach for how to address the forgetting problem via compressing the distributed sensor data to increase the capacity of the CL memory bank using a compressive learning algorithm. Through extensive experiments, we show that our approach provides around 8% accuracy improvement compared to other CL algorithms when applied to a real-world distributed sensor dataset collected from an oilfield. Noticeable accuracy improvement is also achieved when using our proposed approach with the CL benchmark datasets, achieving state-of-the-art accuracies for the CIFAR-10 dataset on blurry10 and blurry30 settings of 80.83% and 88.91%, respectively.},
author = {Arief, Hasan Asy'ari and Thomas, Peter James and Constable, Kevin and Katsaggelos, Aggelos K.},
doi = {10.3390/s22249878},
issn = {1424-8220},
journal = {Sensors},
keywords = {continual learning,distributed acoustic sensing,virtual flow meter},
month = {dec},
number = {24},
pages = {9878},
pmid = {36560248},
publisher = {MDPI},
title = {{Towards Building a Distributed Virtual Flow Meter via Compressed Continual Learning}},
url = {https://www.mdpi.com/1424-8220/22/24/9878},
volume = {22},
year = {2022}
}

A robust–accurate estimation of fluid flow is the main building block of a distributed virtual flow meter. Unfortunately, a big leap in algorithm development would be required for this objective to come to fruition, mainly due to the inability of current machine learning algorithms to make predictions outside the training data distribution. To improve predictions outside the training distribution, we explore the continual learning (CL) paradigm for accurately estimating the characteristics of fluid flow in pipelines. A significant challenge facing CL is the concept of catastrophic forgetting. In this paper, we provide a novel approach for how to address the forgetting problem via compressing the distributed sensor data to increase the capacity of the CL memory bank using a compressive learning algorithm. Through extensive experiments, we show that our approach provides around 8% accuracy improvement compared to other CL algorithms when applied to a real-world distributed sensor dataset collected from an oilfield. Noticeable accuracy improvement is also achieved when using our proposed approach with the CL benchmark datasets, achieving state-of-the-art accuracies for the CIFAR-10 dataset on blurry10 and blurry30 settings of 80.83% and 88.91%, respectively.

@article{KyleAkira2022,
abstract = {Binary stars undergo a variety of interactions and evolutionary phases, critical for predicting and explaining observations. Binary population synthesis with full simulation of stellar structure and evolution is computationally expensive, requiring a large number of mass-transfer sequences. The recently developed binary population synthesis code POSYDON incorporates grids of MESA binary star simulations that are interpolated to model large-scale populations of massive binaries. The traditional method of computing a high-density rectilinear grid of simulations is not scalable for higher-dimension grids, accounting for a range of metallicities, rotation, and eccentricity. We present a new active learning algorithm, psy-cris , which uses machine learning in the data-gathering process to adaptively and iteratively target simulations to run, resulting in a custom, high-performance training set. We test psy-cris on a toy problem and find the resulting training sets require fewer simulations for accurate classification and regression than either regular or randomly sampled grids. We further apply psy-cris to the target problem of building a dynamic grid of MESA simulations, and we demonstrate that, even without fine tuning, a simulation set of only ∼1/4 the size of a rectilinear grid is sufficient to achieve the same classification accuracy. We anticipate further gains when algorithmic parameters are optimized for the targeted application. We find that optimizing for classification only may lead to performance losses in regression, and vice versa. Lowering the computational cost of producing grids will enable new population synthesis codes such as POSYDON to cover more input parameters while preserving interpolation accuracies.},
archivePrefix = {arXiv},
arxivId = {2203.16683},
author = {Rocha, Kyle Akira and Andrews, Jeff J. and Berry, Christopher P. L. and Doctor, Zoheyr and Katsaggelos, Aggelos K and {Serra P{\'{e}}rez}, Juan Gabriel and Marchant, Pablo and Kalogera, Vicky and Coughlin, Scott and Bavera, Simone S. and Dotter, Aaron and Fragos, Tassos and Kovlakas, Konstantinos and Misra, Devina and Xing, Zepei and Zapartas, Emmanouil},
doi = {10.3847/1538-4357/ac8b05},
eprint = {2203.16683},
issn = {0004-637X},
journal = {The Astrophysical Journal},
month = {oct},
number = {1},
pages = {64},
title = {{Active Learning for Computationally Efficient Distribution of Binary Evolution Simulations}},
url = {https://iopscience.iop.org/article/10.3847/1538-4357/ac8b05},
volume = {938},
year = {2022}
}

Binary stars undergo a variety of interactions and evolutionary phases, critical for predicting and explaining observations. Binary population synthesis with full simulation of stellar structure and evolution is computationally expensive, requiring a large number of mass-transfer sequences. The recently developed binary population synthesis code POSYDON incorporates grids of MESA binary star simulations that are interpolated to model large-scale populations of massive binaries. The traditional method of computing a high-density rectilinear grid of simulations is not scalable for higher-dimension grids, accounting for a range of metallicities, rotation, and eccentricity. We present a new active learning algorithm, psy-cris , which uses machine learning in the data-gathering process to adaptively and iteratively target simulations to run, resulting in a custom, high-performance training set. We test psy-cris on a toy problem and find the resulting training sets require fewer simulations for accurate classification and regression than either regular or randomly sampled grids. We further apply psy-cris to the target problem of building a dynamic grid of MESA simulations, and we demonstrate that, even without fine tuning, a simulation set of only ∼1/4 the size of a rectilinear grid is sufficient to achieve the same classification accuracy. We anticipate further gains when algorithmic parameters are optimized for the targeted application. We find that optimizing for classification only may lead to performance losses in regression, and vice versa. Lowering the computational cost of producing grids will enable new population synthesis codes such as POSYDON to cover more input parameters while preserving interpolation accuracies.

@article{Lionel,
abstract = {Optical coherence tomography (OCT) is an optical technique which allows for volumetric visualization of the internal structures of translucent materials. Additional information can be gained by measuring the rate of signal attenuation in depth. Techniques have been developed to estimate the rate of attenuation on a voxel by voxel basis. This depth resolved attenuation analysis gives insight into tissue structure and organization in a spatially resolved way. However, the presence of speckle in the OCT measurement causes the attenuation coefficient image to contain unrealistic fluctuations and makes the reliability of these images at the voxel level poor. While the distribution of speckle in OCT images has appeared in literature, the resulting voxelwise corruption of the attenuation analysis has not. In this work, the estimated depth resolved attenuation coefficient from OCT data with speckle is shown to be approximately exponentially distributed. After this, a prior distribution for the depth resolved attenuation coefficient is derived for a simple system using statistical mechanics. Finally, given a set of depth resolved estimates which were made from OCT data in the presence of speckle, a posterior probability distribution for the true voxelwise attenuation coefficient is derived and a Bayesian voxelwise estimator for the coefficient is given. These results are demonstrated in simulation and validated experimentally.},
author = {Fiske, Lionel D. and Aalders, Maurice C. G. and Almasian, Mitra and van Leeuwen, Ton G. and Katsaggelos, Aggelos K. and Cossairt, Oliver and Faber, Dirk J.},
doi = {10.1038/s41598-021-81713-7},
issn = {2045-2322},
journal = {Scientific Reports},
month = {jan},
number = {1},
pages = {2263},
pmid = {33500435},
title = {{Bayesian analysis of depth resolved OCT attenuation coefficients}},
url = {https://www.nature.com/articles/s41598-021-81713-7},
volume = {11},
year = {2021}
}

Optical coherence tomography (OCT) is an optical technique which allows for volumetric visualization of the internal structures of translucent materials. Additional information can be gained by measuring the rate of signal attenuation in depth. Techniques have been developed to estimate the rate of attenuation on a voxel by voxel basis. This depth resolved attenuation analysis gives insight into tissue structure and organization in a spatially resolved way. However, the presence of speckle in the OCT measurement causes the attenuation coefficient image to contain unrealistic fluctuations and makes the reliability of these images at the voxel level poor. While the distribution of speckle in OCT images has appeared in literature, the resulting voxelwise corruption of the attenuation analysis has not. In this work, the estimated depth resolved attenuation coefficient from OCT data with speckle is shown to be approximately exponentially distributed. After this, a prior distribution for the depth resolved attenuation coefficient is derived for a simple system using statistical mechanics. Finally, given a set of depth resolved estimates which were made from OCT data in the presence of speckle, a posterior probability distribution for the true voxelwise attenuation coefficient is derived and a Bayesian voxelwise estimator for the coefficient is given. These results are demonstrated in simulation and validated experimentally.

@article{Srutarshi2021,
abstract = {We present a novel adaptive host-chip modular architecture for video acquisition to optimize an overall objective task constrained under a given bit rate. The chip is a high resolution imaging sensor such as gigapixel focal plane array (FPA) with low computational power deployed on the field remotely, while the host is a server with high computational power. The communication channel data bandwidth between the chip and host is constrained to accommodate transfer of all captured data from the chip. The host performs objective task specific computations and also intelligently guides the chip to optimize (compress) the data sent to host. This proposed system is modular and highly versatile in terms of flexibility in re-orienting the objective task. In this work, object tracking is the objective task. While our architecture supports any form of compression/distortion, in this paper we use quadtree (QT)-segmented video frames. We use Viterbi (Dynamic Programming) algorithm to minimize the area normalized weighted rate-distortion allocation of resources. The host receives only these degraded frames for analysis. An object detector is used to detect objects, and a Kalman Filter based tracker is used to track those objects. Evaluation of system performance is done in terms of Multiple Object Tracking Accuracy (MOTA) metric. In this proposed novel architecture, performance gains in MOTA is obtained by twice training the object detector with different system generated distortions as a novel 2-step process. Additionally, object detector is assisted by tracker to upscore the region proposals in the detector to further improve the performance.},
archivePrefix = {arXiv},
arxivId = {2102.12046},
author = {Banerjee, Srutarshi and Chopp, Henry H. and Serra, Juan G. and Yang, Hao Tian and Cossairt, Oliver and Katsaggelos, Aggelos K.},
doi = {10.1109/JSEN.2021.3081351},
eprint = {2102.12046},
issn = {1530-437X},
journal = {IEEE Sensors Journal},
keywords = {Image acquisition,Viterbi algorithm,image reconstruction,object detection,object tracker assisted detection,object tracking,optimization,video signal processing},
month = {aug},
number = {15},
pages = {17227--17243},
title = {{An Adaptive Video Acquisition Scheme for Object Tracking and Its Performance Optimization}},
url = {https://ieeexplore.ieee.org/document/9432960/},
volume = {21},
year = {2021}
}

We present a novel adaptive host-chip modular architecture for video acquisition to optimize an overall objective task constrained under a given bit rate. The chip is a high resolution imaging sensor such as gigapixel focal plane array (FPA) with low computational power deployed on the field remotely, while the host is a server with high computational power. The communication channel data bandwidth between the chip and host is constrained to accommodate transfer of all captured data from the chip. The host performs objective task specific computations and also intelligently guides the chip to optimize (compress) the data sent to host. This proposed system is modular and highly versatile in terms of flexibility in re-orienting the objective task. In this work, object tracking is the objective task. While our architecture supports any form of compression/distortion, in this paper we use quadtree (QT)-segmented video frames. We use Viterbi (Dynamic Programming) algorithm to minimize the area normalized weighted rate-distortion allocation of resources. The host receives only these degraded frames for analysis. An object detector is used to detect objects, and a Kalman Filter based tracker is used to track those objects. Evaluation of system performance is done in terms of Multiple Object Tracking Accuracy (MOTA) metric. In this proposed novel architecture, performance gains in MOTA is obtained by twice training the object detector with different system generated distortions as a novel 2-step process. Additionally, object detector is assisted by tracker to upscore the region proposals in the detector to further improve the performance.

@article{Petmezas2021,
abstract = {Atrial fibrillation is a heart arrhythmia strongly associated with other heart-related complications that can increase the risk of strokes and heart failure. Manual electrocardiogram (ECG) interpretation for its diagnosis is tedious, time-consuming, requires high expertise, and suffers from inter- and intra-observer variability. Deep learning techniques could be exploited in order for robust arrhythmia detection models to be designed. In this paper, we propose a novel hybrid neural model utilizing focal loss, an improved version of cross-entropy loss, to deal with training data imbalance. ECG features initially extracted via a Convolutional Neural Network (CNN) are input to a Long Short-Term Memory (LSTM) model for temporal dynamics memorization and thus, more accurate classification into the four ECG rhythm types, namely normal (N), atrial fibrillation (AFIB), atrial flutter (AFL) and AV junctional rhythm (J). The model was trained on the MIT-BIH Atrial Fibrillation Database and achieved a sensitivity of 97.87%, and specificity of 99.29% using a ten-fold cross-validation strategy. The proposed model can aid clinicians to detect common atrial fibrillation in real-time on routine screening ECG.},
author = {Petmezas, Georgios and Haris, Kostas and Stefanopoulos, Leandros and Kilintzis, Vassilis and Tzavelis, Andreas and Rogers, John A. and Katsaggelos, Aggelos K. and Maglaveras, Nicos},
doi = {10.1016/j.bspc.2020.102194},
issn = {17468094},
journal = {Biomedical Signal Processing and Control},
keywords = {CNN,LSTM,arrhythmia detection,atrial fibrillation,focal loss},
month = {jan},
pages = {102194},
publisher = {Elsevier},
title = {{Automated Atrial Fibrillation Detection using a Hybrid CNN-LSTM Network on Imbalanced ECG Datasets}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1746809420303323},
volume = {63},
year = {2021}
}

Atrial fibrillation is a heart arrhythmia strongly associated with other heart-related complications that can increase the risk of strokes and heart failure. Manual electrocardiogram (ECG) interpretation for its diagnosis is tedious, time-consuming, requires high expertise, and suffers from inter- and intra-observer variability. Deep learning techniques could be exploited in order for robust arrhythmia detection models to be designed. In this paper, we propose a novel hybrid neural model utilizing focal loss, an improved version of cross-entropy loss, to deal with training data imbalance. ECG features initially extracted via a Convolutional Neural Network (CNN) are input to a Long Short-Term Memory (LSTM) model for temporal dynamics memorization and thus, more accurate classification into the four ECG rhythm types, namely normal (N), atrial fibrillation (AFIB), atrial flutter (AFL) and AV junctional rhythm (J). The model was trained on the MIT-BIH Atrial Fibrillation Database and achieved a sensitivity of 97.87%, and specificity of 99.29% using a ten-fold cross-validation strategy. The proposed model can aid clinicians to detect common atrial fibrillation in real-time on routine screening ECG.

@article{Wehbe2021,
abstract = {Background: There are characteristic findings of coronavirus disease 2019 (COVID-19) on chest images. An artificial intelligence (AI) algorithm to detect COVID-19 on chest radiographs might be useful for triage or infection control within a hospital setting, but prior reports have been limited by small data sets, poor data quality, or both. Purpose: To present DeepCOVID-XR, a deep learning AI algorithm to detect COVID-19 on chest radiographs, that was trained and tested on a large clinical data set. Materials and Methods: DeepCOVID-XR is an ensemble of convolutional neural networks developed to detect COVID-19 on frontal chest radiographs, with reverse-transcription polymerase chain reaction test results as the reference standard. The algorithm was trained and validated on 14 788 images (4253 positive for COVID-19) from sites across the Northwestern Memorial Health Care System from February 2020 to April 2020 and was then tested on 2214 images (1192 positive for COVID-19) from a single holdout institution. Performance of the algorithm was compared with interpretations from five experienced thoracic radiologists on 300 random test images using the McNemar test for sensitivity and specificity and the DeLong test for the area under the receiver operating characteristic curve (AUC). Results: A total of 5853 patients (mean age, 58 years 6 19 [standard deviation]; 3101 women) were evaluated across data sets. For the entire test set, accuracy of DeepCOVID-XR was 83%, with an AUC of 0.90. For 300 random test images (134 positive for COVID-19), accuracy of DeepCOVID-XR was 82%, compared with that of individual radiologists (range, 76%–81%) and the consensus of all five radiologists (81%). DeepCOVID-XR had a significantly higher sensitivity (71%) than one radiologist (60%, P , .001) and significantly higher specificity (92%) than two radiologists (75%, P , .001; 84%, P = .009). AUC of DeepCOVID-XR was 0.88 compared with the consensus AUC of 0.85 (P = .13 for comparison). With consensus interpretation as the reference standard, the AUC of DeepCOVID-XR was 0.95 (95% CI: 0.92, 0.98). Conclusion: DeepCOVID-XR, an artificial intelligence algorithm, detected coronavirus disease 2019 on chest radiographs with a performance similar to that of experienced thoracic radiologists in consensus.},
annote = {doi: 10.1148/radiol.2020203511},
author = {Wehbe, Ramsey M. and Sheng, Jiayue and Dutta, Shinjan and Chai, Siyuan and Dravid, Amil and Barutcu, Semih and Wu, Yunan and Cantrell, Donald R. and Xiao, Nicholas and Allen, Bradley D. and MacNealy, Gregory A. and Savas, Hatice and Agrawal, Rishi and Parekh, Nishant and Katsaggelos, Aggelos K.},
doi = {10.1148/radiol.2020203511},
issn = {0033-8419},
journal = {Radiology},
month = {apr},
number = {1},
pages = {E167--E176},
pmid = {33231531},
publisher = {Radiological Society of North America},
title = {{DeepCOVID-XR: An Artificial Intelligence Algorithm to Detect COVID-19 on Chest Radiographs Trained and Tested on a Large U.S. Clinical Data Set}},
url = {https://doi.org/10.1148/radiol.2020203511 http://pubs.rsna.org/doi/10.1148/radiol.2020203511},
volume = {299},
year = {2021}
}

Background: There are characteristic findings of coronavirus disease 2019 (COVID-19) on chest images. An artificial intelligence (AI) algorithm to detect COVID-19 on chest radiographs might be useful for triage or infection control within a hospital setting, but prior reports have been limited by small data sets, poor data quality, or both. Purpose: To present DeepCOVID-XR, a deep learning AI algorithm to detect COVID-19 on chest radiographs, that was trained and tested on a large clinical data set. Materials and Methods: DeepCOVID-XR is an ensemble of convolutional neural networks developed to detect COVID-19 on frontal chest radiographs, with reverse-transcription polymerase chain reaction test results as the reference standard. The algorithm was trained and validated on 14 788 images (4253 positive for COVID-19) from sites across the Northwestern Memorial Health Care System from February 2020 to April 2020 and was then tested on 2214 images (1192 positive for COVID-19) from a single holdout institution. Performance of the algorithm was compared with interpretations from five experienced thoracic radiologists on 300 random test images using the McNemar test for sensitivity and specificity and the DeLong test for the area under the receiver operating characteristic curve (AUC). Results: A total of 5853 patients (mean age, 58 years 6 19 [standard deviation]; 3101 women) were evaluated across data sets. For the entire test set, accuracy of DeepCOVID-XR was 83%, with an AUC of 0.90. For 300 random test images (134 positive for COVID-19), accuracy of DeepCOVID-XR was 82%, compared with that of individual radiologists (range, 76%–81%) and the consensus of all five radiologists (81%). DeepCOVID-XR had a significantly higher sensitivity (71%) than one radiologist (60%, P , .001) and significantly higher specificity (92%) than two radiologists (75%, P , .001; 84%, P = .009). AUC of DeepCOVID-XR was 0.88 compared with the consensus AUC of 0.85 (P = .13 for comparison). With consensus interpretation as the reference standard, the AUC of DeepCOVID-XR was 0.95 (95% CI: 0.92, 0.98). Conclusion: DeepCOVID-XR, an artificial intelligence algorithm, detected coronavirus disease 2019 on chest radiographs with a performance similar to that of experienced thoracic radiologists in consensus.

@article{besson2021geometric,
abstract = {The complex relationship between the shape and function of the human brain remains elusive despite extensive studies of cortical folding over many decades. The analysis of cortical gyrification presents an opportunity to advance our knowledge about this relationship, and better understand the etiology of a variety of pathologies involving diverse degrees of cortical folding abnormalities. Hypothesis-driven surface-based approaches have been shown to be particularly efficient in their ability to accurately describe unique features of the folded sheet topology of the cortical ribbon. However, the utility of these approaches has been blunted by their reliance on manually defined features aiming to capture the relevant geometric properties of cortical folding. In this paper, we propose an entirely novel, data-driven deep-learning based method to analyze the brain's shape that eliminates this reliance on manual feature definition. This method builds on the emerging field of geometric deep-learning and uses traditional convolutional neural network architecture uniquely adapted to the surface representation of the cortical ribbon. This method is a complete departure from prior brain MRI CNN investigations, all of which have relied on three dimensional MRI data and interpreted features of the MRI signal for prediction. MRI data from 6410 healthy subjects obtained from 11 publicly available data repositories were used for analysis. Ages ranged from 6 to 89 years. Both inner and outer cortical surfaces were extracted using Freesurfer and then registered into MNI space. For purposes of method development, both a classification and regression challenge were introduced for network learning including sex and age prediction, respectively. Two independent graph convolutional neural networks (gCNNs) were trained, the first of which to predict subject's self-identified sex, the second of which to predict subject's age. Class Activation Maps (CAM) and Regression Activation Maps (RAM) were constructed respectively to map the topographic distribution of the most influential brain regions involved in the decision process for each gCNN. Using this approach, the gCNN was able to predict a subject's sex with an average accuracy of 87.99 % and achieved a Person's coefficient of correlation of 0.93 with an average absolute error 4.58 years when predicting a subject's age. We believe this shape-based convolutional classifier offers a novel, data-driven approach to define biomedically relevant features from the brain at both the population and single subject levels and therefore lays a critical foundation for future precision medicine applications.},
author = {Besson, Pierre and Parrish, Todd and Katsaggelos, Aggelos K. and Bandt, S. Kathleen},
doi = {10.1016/j.compmedimag.2021.101939},
issn = {08956111},
journal = {Computerized Medical Imaging and Graphics},
keywords = {Big data,Brain mapping,Brain shape,Geometric deep learning,Population health,Precision medicine},
month = {jul},
pages = {101939},
pmid = {34082280},
publisher = {Pergamon},
title = {{Geometric deep learning on brain shape predicts sex and age}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0895611121000884},
volume = {91},
year = {2021}
}

The complex relationship between the shape and function of the human brain remains elusive despite extensive studies of cortical folding over many decades. The analysis of cortical gyrification presents an opportunity to advance our knowledge about this relationship, and better understand the etiology of a variety of pathologies involving diverse degrees of cortical folding abnormalities. Hypothesis-driven surface-based approaches have been shown to be particularly efficient in their ability to accurately describe unique features of the folded sheet topology of the cortical ribbon. However, the utility of these approaches has been blunted by their reliance on manually defined features aiming to capture the relevant geometric properties of cortical folding. In this paper, we propose an entirely novel, data-driven deep-learning based method to analyze the brain's shape that eliminates this reliance on manual feature definition. This method builds on the emerging field of geometric deep-learning and uses traditional convolutional neural network architecture uniquely adapted to the surface representation of the cortical ribbon. This method is a complete departure from prior brain MRI CNN investigations, all of which have relied on three dimensional MRI data and interpreted features of the MRI signal for prediction. MRI data from 6410 healthy subjects obtained from 11 publicly available data repositories were used for analysis. Ages ranged from 6 to 89 years. Both inner and outer cortical surfaces were extracted using Freesurfer and then registered into MNI space. For purposes of method development, both a classification and regression challenge were introduced for network learning including sex and age prediction, respectively. Two independent graph convolutional neural networks (gCNNs) were trained, the first of which to predict subject's self-identified sex, the second of which to predict subject's age. Class Activation Maps (CAM) and Regression Activation Maps (RAM) were constructed respectively to map the topographic distribution of the most influential brain regions involved in the decision process for each gCNN. Using this approach, the gCNN was able to predict a subject's sex with an average accuracy of 87.99 % and achieved a Person's coefficient of correlation of 0.93 with an average absolute error 4.58 years when predicting a subject's age. We believe this shape-based convolutional classifier offers a novel, data-driven approach to define biomedically relevant features from the brain at both the population and single subject levels and therefore lays a critical foundation for future precision medicine applications.

@article{Henry2021,
abstract = {In this paper, we propose EveRestNet, a convolutional neural network designed to remove blocking artifacts in videostreams using events from neuromorphic sensors. We first degrade the video frame using a quadtree structure to produce the blocking artifacts to simulate transmitting a video under a heavily constrained bandwidth. Events from the neuromorphic sensor are also simulated, but are transmitted in full. Using the distorted frames and the event stream, EveRestNet is able to improve the image quality.},
archivePrefix = {arXiv},
arxivId = {2105.05973},
author = {Chopp, Henry H. and Banerjee, Srutarshi and Cossairt, Oliver and Katsaggelos, Aggelos K.},
eprint = {2105.05973},
journal = {arXiv preprint arXiv:2105.05973},
month = {may},
title = {{Removing Blocking Artifacts in Video Streams Using Event Cameras}},
url = {http://arxiv.org/abs/2105.05973},
year = {2021}
}

In this paper, we propose EveRestNet, a convolutional neural network designed to remove blocking artifacts in videostreams using events from neuromorphic sensors. We first degrade the video frame using a quadtree structure to produce the blocking artifacts to simulate transmitting a video under a heavily constrained bandwidth. Events from the neuromorphic sensor are also simulated, but are transmitted in full. Using the distorted frames and the event stream, EveRestNet is able to improve the image quality.

@article{Siddharth2021,
abstract = {The observation of gravitational waves is hindered by the presence of transient noise (glitches). We study data from the third observing run of the Advanced LIGO detectors, and identify new glitch classes: fast scattering/crown and low-frequency blips. Using training sets assembled by monitoring of the state of the detector, and by citizen-science volunteers, we update the Gravity Spy machine-learning algorithm for glitch classification. We find that fast scattering/crown, linked to ground motion at the detector sites, is especially prevalent, and identify two subclasses linked to different types of ground motion. Reclassification of data based on the updated model finds that ∼27% of all transient noise at LIGO Livingston belongs to the fast scattering class, while ∼8% belongs to the low-frequency blip class, making them the most frequent and fourth most frequent sources of transient noise at that site. Our results demonstrate both how glitch classification can reveal potential improvements to gravitational-wave detectors, and how, given an appropriate framework, citizen-science volunteers may make discoveries in large data sets.},
archivePrefix = {arXiv},
arxivId = {2103.12104},
author = {Soni, S. and Berry, C P L and Coughlin, S. B. and Harandi, M. and Jackson, C. B. and Crowston, K. and {\O}sterlund, C and Patane, O. and Katsaggelos, A. K. and Trouille, L. and Baranowski, V-G and Domainko, W. F. and Kaminski, K. and Rodriguez, M A Lobato and Marciniak, U. and Nauta, P. and Niklasch, G. and Rote, R. R. and T{\'{e}}gl{\'{a}}s, B. and Unsworth, C. and Zhang, C.},
doi = {10.1088/1361-6382/ac1ccb},
eprint = {2103.12104},
issn = {0264-9381},
journal = {Classical and Quantum Gravity},
keywords = {LIGO,machine learning,neural network,noise classification,transient noise},
month = {oct},
number = {19},
pages = {195016},
title = {{Discovering features in gravitational-wave data through detector characterization, citizen science and machine learning}},
url = {https://iopscience.iop.org/article/10.1088/1361-6382/ac1ccb},
volume = {38},
year = {2021}
}

The observation of gravitational waves is hindered by the presence of transient noise (glitches). We study data from the third observing run of the Advanced LIGO detectors, and identify new glitch classes: fast scattering/crown and low-frequency blips. Using training sets assembled by monitoring of the state of the detector, and by citizen-science volunteers, we update the Gravity Spy machine-learning algorithm for glitch classification. We find that fast scattering/crown, linked to ground motion at the detector sites, is especially prevalent, and identify two subclasses linked to different types of ground motion. Reclassification of data based on the updated model finds that ∼27% of all transient noise at LIGO Livingston belongs to the fast scattering class, while ∼8% belongs to the low-frequency blip class, making them the most frequent and fourth most frequent sources of transient noise at that site. Our results demonstrate both how glitch classification can reveal potential improvements to gravitational-wave detectors, and how, given an appropriate framework, citizen-science volunteers may make discoveries in large data sets.

@article{Peiqi2021,
abstract = {Many visual and robotics tasks in real-world scenarios rely on robust handling of high speed motion and high dynamic range (HDR) with effectively high spatial resolution and low noise. Such stringent requirements, however, cannot be directly satisfied by a single imager or imaging modality, rather by multi-modal sensors with complementary advantages. In this paper, we address high performance imaging by exploring the synergy between traditional frame-based sensors with high spatial resolution and low sensor noise, and emerging event-based sensors with high speed and high dynamic range. We introduce a novel computational framework, termed Guided Event Filtering (GEF), to process these two streams of input data and output a stream of super-resolved yet noise-reduced events. To generate high quality events, GEF first registers the captured noisy events onto the guidance image plane according to our flow model. it then performs joint image filtering that inherits the mutual structure from both inputs. Lastly, GEF re-distributes the filtered event frame in the space-time volume while preserving the statistical characteristics of the original events. When the guidance images under-perform, GEF incorporates an event self-guiding mechanism that resorts to neighbor events for guidance. We demonstrate the benefits of GEF by applying the output high quality events to existing event-based algorithms across diverse application categories, including high speed object tracking, depth estimation, high frame-rate video synthesis, and super resolution/HDR/color image restoration.},
author = {Duan, Peiqi and Wang, Zihao and Shi, Boxin and Cossairt, Oliver and Huang, Tiejun and Katsaggelos, Aggelos},
doi = {10.1109/TPAMI.2021.3113344},
issn = {0162-8828},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
keywords = {Computational hybrid cameras,event-based imaging and vision,joint filtering},
number = {11},
pages = {1--1},
pmid = {34543190},
title = {{Guided Event Filtering: Synergy between Intensity Images and Neuromorphic Events for High Performance Imaging}},
url = {https://ieeexplore.ieee.org/document/9541050/},
volume = {44},
year = {2021}
}

Many visual and robotics tasks in real-world scenarios rely on robust handling of high speed motion and high dynamic range (HDR) with effectively high spatial resolution and low noise. Such stringent requirements, however, cannot be directly satisfied by a single imager or imaging modality, rather by multi-modal sensors with complementary advantages. In this paper, we address high performance imaging by exploring the synergy between traditional frame-based sensors with high spatial resolution and low sensor noise, and emerging event-based sensors with high speed and high dynamic range. We introduce a novel computational framework, termed Guided Event Filtering (GEF), to process these two streams of input data and output a stream of super-resolved yet noise-reduced events. To generate high quality events, GEF first registers the captured noisy events onto the guidance image plane according to our flow model. it then performs joint image filtering that inherits the mutual structure from both inputs. Lastly, GEF re-distributes the filtered event frame in the space-time volume while preserving the statistical characteristics of the original events. When the guidance images under-perform, GEF incorporates an event self-guiding mechanism that resorts to neighbor events for guidance. We demonstrate the benefits of GEF by applying the output high quality events to existing event-based algorithms across diverse application categories, including high speed object tracking, depth estimation, high frame-rate video synthesis, and super resolution/HDR/color image restoration.

@article{bae2019transfer,
abstract = {Dementia of Alzheimer's type (DAT) is associated with devastating and irreversible cognitive decline. Predicting which patients with mild cognitive impairment (MCI) will progress to DAT is an ongoing challenge in the field. We developed a deep learning model to predict conversion from MCI to DAT. Structural magnetic resonance imaging scans were used as input to a 3-dimensional convolutional neural network. The 3-dimensional convolutional neural network was trained using transfer learning; in the source task, normal control and DAT scans were used to pretrain the model. This pretrained model was then retrained on the target task of classifying which MCI patients converted to DAT. Our model resulted in 82.4% classification accuracy at the target task, outperforming current models in the field. Next, we visualized brain regions that significantly contribute to the prediction of MCI conversion using an occlusion map approach. Contributory regions included the pons, amygdala, and hippocampus. Finally, we showed that the model's prediction value is significantly correlated with rates of change in clinical assessment scores, indicating that the model is able to predict an individual patient's future cognitive decline. This information, in conjunction with the identified anatomical features, will aid in building a personalized therapeutic strategy for individuals with MCI.},
author = {Bae, Jinhyeong and Stocks, Jane and Heywood, Ashley and Jung, Youngmoon and Jenkins, Lisanne and Hill, Virginia and Katsaggelos, Aggelos and Popuri, Karteek and Rosen, Howie and Beg, M. Faisal and Wang, Lei},
doi = {10.1016/j.neurobiolaging.2020.12.005},
issn = {01974580},
journal = {Neurobiology of Aging},
keywords = {Convolutional neural network,Dementia of Alzheimer's type,Magnetic resonance imaging,Mild cognitive impairment,Predictive modeling},
month = {mar},
pages = {53--64},
pmid = {33422894},
publisher = {Cold Spring Harbor Laboratory},
title = {{Transfer learning for predicting conversion from mild cognitive impairment to dementia of Alzheimer's type based on a three-dimensional convolutional neural network}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0197458020304140},
volume = {99},
year = {2021}
}

Dementia of Alzheimer's type (DAT) is associated with devastating and irreversible cognitive decline. Predicting which patients with mild cognitive impairment (MCI) will progress to DAT is an ongoing challenge in the field. We developed a deep learning model to predict conversion from MCI to DAT. Structural magnetic resonance imaging scans were used as input to a 3-dimensional convolutional neural network. The 3-dimensional convolutional neural network was trained using transfer learning; in the source task, normal control and DAT scans were used to pretrain the model. This pretrained model was then retrained on the target task of classifying which MCI patients converted to DAT. Our model resulted in 82.4% classification accuracy at the target task, outperforming current models in the field. Next, we visualized brain regions that significantly contribute to the prediction of MCI conversion using an occlusion map approach. Contributory regions included the pons, amygdala, and hippocampus. Finally, we showed that the model's prediction value is significantly correlated with rates of change in clinical assessment scores, indicating that the model is able to predict an individual patient's future cognitive decline. This information, in conjunction with the identified anatomical features, will aid in building a personalized therapeutic strategy for individuals with MCI.

@article{Fernando2021,
abstract = {Background and Objective:Color variations in digital histopathology severely impact the performance of computer-aided diagnosis systems. They are due to differences in the staining process and acquisition system, among other reasons. Blind color deconvolution techniques separate multi-stained images into single stained bands which, once normalized, can be used to eliminate these negative color variations and improve the performance of machine learning tasks. Methods:In this work, we decompose the observed RGB image in its hematoxylin and eosin components. We apply Bayesian modeling and inference based on the use of Super Gaussian sparse priors for each stain together with prior closeness to a given reference color-vector matrix. The hematoxylin and eosin components are then used for image normalization and classification of histological images. The proposed framework is tested on stain separation, image normalization, and cancer classification problems. The results are measured using the peak signal to noise ratio, normalized median intensity and the area under ROC curve on five different databases. Results:The obtained results show the superiority of our approach to current state-of-the-art blind color deconvolution techniques. In particular, the fidelity to the tissue improves 1,27 dB in mean PSNR. The normalized median intensity shows a good normalization quality of the proposed approach on the tested datasets. Finally, in cancer classification experiments the area under the ROC curve improves from 0.9491 to 0.9656 and from 0.9279 to 0.9541 on Camelyon-16 and Camelyon-17, respectively, when the original and processed images are used. Furthermore, these figures of merits are better than those obtained by the methods compared with. Conclusions:The proposed framework for blind color deconvolution, normalization and classification of images guarantees fidelity to the tissue structure and can be used both for normalization and classification. In addition, color deconvolution enables the use of the optical density space for classification, which improves the classification performance.},
author = {P{\'{e}}rez-Bueno, Fernando and Vega, Miguel and Sales, Mar{\'{i}}a A. and Aneiros-Fern{\'{a}}ndez, Jos{\'{e}} and Naranjo, Valery and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.cmpb.2021.106453},
issn = {01692607},
journal = {Computer Methods and Programs in Biomedicine},
keywords = {Blind color deconvolution,Histopathological images,Image normalization,Super Gaussian,Variational bayes},
month = {nov},
pages = {106453},
pmid = {34649072},
title = {{Blind color deconvolution, normalization, and classification of histological images using general super Gaussian priors and Bayesian inference}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0169260721005277},
volume = {211},
year = {2021}
}

Background and Objective:Color variations in digital histopathology severely impact the performance of computer-aided diagnosis systems. They are due to differences in the staining process and acquisition system, among other reasons. Blind color deconvolution techniques separate multi-stained images into single stained bands which, once normalized, can be used to eliminate these negative color variations and improve the performance of machine learning tasks. Methods:In this work, we decompose the observed RGB image in its hematoxylin and eosin components. We apply Bayesian modeling and inference based on the use of Super Gaussian sparse priors for each stain together with prior closeness to a given reference color-vector matrix. The hematoxylin and eosin components are then used for image normalization and classification of histological images. The proposed framework is tested on stain separation, image normalization, and cancer classification problems. The results are measured using the peak signal to noise ratio, normalized median intensity and the area under ROC curve on five different databases. Results:The obtained results show the superiority of our approach to current state-of-the-art blind color deconvolution techniques. In particular, the fidelity to the tissue improves 1,27 dB in mean PSNR. The normalized median intensity shows a good normalization quality of the proposed approach on the tested datasets. Finally, in cancer classification experiments the area under the ROC curve improves from 0.9491 to 0.9656 and from 0.9279 to 0.9541 on Camelyon-16 and Camelyon-17, respectively, when the original and processed images are used. Furthermore, these figures of merits are better than those obtained by the methods compared with. Conclusions:The proposed framework for blind color deconvolution, normalization and classification of images guarantees fidelity to the tissue structure and can be used both for normalization and classification. In addition, color deconvolution enables the use of the optical density space for classification, which improves the classification performance.

@article{cheimariotis2021automatic,
abstract = {Intravascular Optical Coherence Tomography (IVOCT) images provide important insight into every aspect of atherosclerosis. Specifically, the extent of plaque and its type, which are indicative of the patient's condition, are better assessed by OCT images in comparison to other in vivo modalities. A large amount of imaging data per patient require automatic methods for rapid results. An effective step towards automatic plaque detection and plaque characterization is axial lines (A-lines) based classification into normal and various plaque types. In this work, a novel automatic method for A-line classification is proposed. The method employed convolutional neural networks (CNNs) for classification in its core and comprised the following pre-processing steps: arterial wall segmentation and an OCT-specific (depth-resolved) transformation and a post-processing step based on the majority of classifications. The important step was the OCT-specific transformation, which was based on the estimation of the attenuation coefficient in every pixel of the OCT image. The dataset used for training and testing consisted of 183 images from 33 patients. In these images, four different plaque types were delineated. The method was evaluated by cross-validation. The mean values of accuracy, sensitivity and specificity were 74.73%, 87.78%, and 61.45%, respectively, when classifying into plaque and normal A-lines. When plaque A-lines were classified into fibrolipidic and fibrocalcific, the overall accuracy was 83.47% for A-lines of OCT-specific transformed images and 74.94% for A-lines of original images. This large improvement in accuracy indicates the advantage of using attenuation coefficients when characterizing plaque types. The proposed automatic deep-learning pipeline constitutes a positive contribution to the accurate classification of A-lines in intravascular OCT images.},
author = {Cheimariotis, Grigorios-Aris and Riga, Maria and Haris, Kostas and Toutouzas, Konstantinos and Katsaggelos, Aggelos K. and Maglaveras, Nicos},
doi = {10.3390/app11167412},
issn = {2076-3417},
journal = {Applied Sciences},
keywords = {Atheromatic plaque,CNN,Classification,Deep learning,Intravascular optical coherence tomography},
month = {aug},
number = {16},
pages = {7412},
publisher = {MDPI},
title = {{Automatic Classification of A-Lines in Intravascular OCT Images Using Deep Learning and Estimation of Attenuation Coefficients}},
url = {https://www.mdpi.com/2076-3417/11/16/7412},
volume = {11},
year = {2021}
}

Intravascular Optical Coherence Tomography (IVOCT) images provide important insight into every aspect of atherosclerosis. Specifically, the extent of plaque and its type, which are indicative of the patient's condition, are better assessed by OCT images in comparison to other in vivo modalities. A large amount of imaging data per patient require automatic methods for rapid results. An effective step towards automatic plaque detection and plaque characterization is axial lines (A-lines) based classification into normal and various plaque types. In this work, a novel automatic method for A-line classification is proposed. The method employed convolutional neural networks (CNNs) for classification in its core and comprised the following pre-processing steps: arterial wall segmentation and an OCT-specific (depth-resolved) transformation and a post-processing step based on the majority of classifications. The important step was the OCT-specific transformation, which was based on the estimation of the attenuation coefficient in every pixel of the OCT image. The dataset used for training and testing consisted of 183 images from 33 patients. In these images, four different plaque types were delineated. The method was evaluated by cross-validation. The mean values of accuracy, sensitivity and specificity were 74.73%, 87.78%, and 61.45%, respectively, when classifying into plaque and normal A-lines. When plaque A-lines were classified into fibrolipidic and fibrocalcific, the overall accuracy was 83.47% for A-lines of OCT-specific transformed images and 74.94% for A-lines of original images. This large improvement in accuracy indicates the advantage of using attenuation coefficients when characterizing plaque types. The proposed automatic deep-learning pipeline constitutes a positive contribution to the accurate classification of A-lines in intravascular OCT images.

@article{Srutarshi2021a,
abstract = {We present a novel adaptive multi-modal intensity-event algorithm to optimize an overall objective of object tracking under bit rate constraints for a host-chip architecture. The chip is a computationally resource constrained device acquiring high resolution intensity frames and events, while the host is capable of performing computationally expensive tasks. We develop a joint intensity-neuromorphic event rate-distortion compression framework with a quadtree (QT) based compression of intensity and events scheme. The data acquisition on the chip is driven by the presence of objects of interest in the scene as detected by an object detector. The most informative intensity and event data are communicated to the host under rate constraints, so that the best possible tracking performance is obtained. The detection and tracking of objects in the scene are done on the distorted data at the host. Intensity and events are jointly used in a fusion framework to enhance the quality of the distorted images, so as to improve the object detection and tracking performance. The performance assessment of the overall system is done in terms of the multiple object tracking accuracy (MOTA) score. Compared to using intensity modality only, there is an improvement in MOTA using both these modalities in different scenarios.},
archivePrefix = {arXiv},
arxivId = {2105.14164},
author = {Banerjee, Srutarshi and Chopp, Henry H and Zhang, Jianping and Wang, Zihao W and Cossairt, Oliver and Katsaggelos, Aggelos},
eprint = {2105.14164},
isbn = {2105.14164v1},
journal = {arXiv preprint arXiv:2105.14164},
keywords = {Compressed domain object detection and tracking,Dynamic Program-ming,Index Terms-Joint Intensity-Event Imaging System,Joint Intensity-Event Rate Distortion Optimization},
month = {may},
title = {{A Joint Intensity-Neuromorphic Event Imaging System for Resource Constrained Devices}},
url = {http://arxiv.org/abs/2105.14164},
year = {2021}
}

We present a novel adaptive multi-modal intensity-event algorithm to optimize an overall objective of object tracking under bit rate constraints for a host-chip architecture. The chip is a computationally resource constrained device acquiring high resolution intensity frames and events, while the host is capable of performing computationally expensive tasks. We develop a joint intensity-neuromorphic event rate-distortion compression framework with a quadtree (QT) based compression of intensity and events scheme. The data acquisition on the chip is driven by the presence of objects of interest in the scene as detected by an object detector. The most informative intensity and event data are communicated to the host under rate constraints, so that the best possible tracking performance is obtained. The detection and tracking of objects in the scene are done on the distorted data at the host. Intensity and events are jointly used in a fusion framework to enhance the quality of the distorted images, so as to improve the object detection and tracking performance. The performance assessment of the overall system is done in terms of the multiple object tracking accuracy (MOTA) score. Compared to using intensity modality only, there is an improvement in MOTA using both these modalities in different scenarios.

@article{Daming2021,
abstract = {Highly accelerated real-time cine MRI using compressed sensing (CS) is a promising approach to achieve high spatio-temporal resolution and clinically acceptable image quality in patients with arrhythmia and/or dyspnea. However, its lengthy image reconstruction time may hinder its clinical translation. The purpose of this study was to develop a neural network for reconstruction of non-Cartesian real-time cine MRI k-space data faster (<1 min per slice with 80 frames) than graphics processing unit (GPU)-accelerated CS reconstruction, without significant loss in image quality or accuracy in left ventricular (LV) functional parameters. We introduce a perceptual complex neural network (PCNN) that trains on complex-valued MRI signal and incorporates a perceptual loss term to suppress incoherent image details. This PCNN was trained and tested with multi-slice, multi-phase, cine images from 40 patients (20 for training, 20 for testing), where the zero-filled images were used as input and the corresponding CS reconstructed images were used as practical ground truth. The resulting images were compared using quantitative metrics (structural similarity index (SSIM) and normalized root mean square error (NRMSE)) and visual scores (conspicuity, temporal fidelity, artifacts, and noise scores), individually graded on a five-point scale (1, worst; 3, acceptable; 5, best), and LV ejection fraction (LVEF). The mean processing time per slice with 80 frames for PCNN was 23.7 ± 1.9 s for pre-processing (Step 1, same as CS) and 0.822 ± 0.004 s for dealiasing (Step 2, 166 times faster than CS). Our PCNN produced higher data fidelity metrics (SSIM = 0.88 ± 0.02, NRMSE = 0.014 ± 0.004) compared with CS. While all the visual scores were significantly different (P < 0.05), the median scores were all 4.0 or higher for both CS and PCNN. LVEFs measured from CS and PCNN were strongly correlated (R2 = 0.92) and in good agreement (mean difference = −1.4% [2.3% of mean]; limit of agreement = 10.6% [17.6% of mean]). The proposed PCNN is capable of rapid reconstruction (25 s per slice with 80 frames) of non-Cartesian real-time cine MRI k-space data, without significant loss in image quality or accuracy in LV functional parameters.},
author = {Shen, Daming and Ghosh, Sushobhan and Haji‐Valizadeh, Hassan and Pathrose, Ashitha and Schiffers, Florian and Lee, Daniel C. and Freed, Benjamin H. and Markl, Michael and Cossairt, Oliver S. and Katsaggelos, Aggelos K. and Kim, Daniel},
doi = {10.1002/nbm.4405},
issn = {0952-3480},
journal = {NMR in Biomedicine},
keywords = {compressed sensing (CS),convolutional neural network (CNN),deep learning (DL),perceptual complex neural network (PCNN),perceptual loss,real-time cine MRI},
month = {jan},
number = {1},
pages = {e4405},
pmid = {32875668},
title = {{Rapid reconstruction of highly undersampled, non‐Cartesian real‐time cine k ‐space data using a perceptual complex neural network (PCNN)}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/nbm.4405},
volume = {34},
year = {2021}
}

Highly accelerated real-time cine MRI using compressed sensing (CS) is a promising approach to achieve high spatio-temporal resolution and clinically acceptable image quality in patients with arrhythmia and/or dyspnea. However, its lengthy image reconstruction time may hinder its clinical translation. The purpose of this study was to develop a neural network for reconstruction of non-Cartesian real-time cine MRI k-space data faster (<1 min per slice with 80 frames) than graphics processing unit (GPU)-accelerated CS reconstruction, without significant loss in image quality or accuracy in left ventricular (LV) functional parameters. We introduce a perceptual complex neural network (PCNN) that trains on complex-valued MRI signal and incorporates a perceptual loss term to suppress incoherent image details. This PCNN was trained and tested with multi-slice, multi-phase, cine images from 40 patients (20 for training, 20 for testing), where the zero-filled images were used as input and the corresponding CS reconstructed images were used as practical ground truth. The resulting images were compared using quantitative metrics (structural similarity index (SSIM) and normalized root mean square error (NRMSE)) and visual scores (conspicuity, temporal fidelity, artifacts, and noise scores), individually graded on a five-point scale (1, worst; 3, acceptable; 5, best), and LV ejection fraction (LVEF). The mean processing time per slice with 80 frames for PCNN was 23.7 ± 1.9 s for pre-processing (Step 1, same as CS) and 0.822 ± 0.004 s for dealiasing (Step 2, 166 times faster than CS). Our PCNN produced higher data fidelity metrics (SSIM = 0.88 ± 0.02, NRMSE = 0.014 ± 0.004) compared with CS. While all the visual scores were significantly different (P < 0.05), the median scores were all 4.0 or higher for both CS and PCNN. LVEFs measured from CS and PCNN were strongly correlated (R2 = 0.92) and in good agreement (mean difference = −1.4% [2.3% of mean]; limit of agreement = 10.6% [17.6% of mean]). The proposed PCNN is capable of rapid reconstruction (25 s per slice with 80 frames) of non-Cartesian real-time cine MRI k-space data, without significant loss in image quality or accuracy in LV functional parameters.

@article{Emanuel2021,
abstract = {Several patterns of atrophy have been identified and strongly related to Alzheimer's disease (AD) pathology and its progression. Morphological changes in brain shape have been identified up to ten years before clinical diagnoses of AD, making its early detection more relevant. We propose novel geometric deep learning frameworks for the analysis of brain shape in the context of neurodegeneration caused by AD. Our deep neural networks learn low-dimensional shape descriptors of multiple neuroanatomical structures, instead of handcrafted features for each structure. A discriminative network using spiral convolution on 3D meshes is constructed for the in-vivo binary classification of AD from healthy controls (HCs) using a fast and efficient "spiral" convolution operator on 3D triangular mesh surfaces of human brain subcortical structures extracted from T1-weighted magnetic resonance imaging (MRI). Our network architecture consists of modular learning blocks using residual connections to improve overall classi-fier performance. In this work: (1) a discriminative network is used to analyze the efficacy of disease classification using input data from multiple brain structures and compared to using a single hemisphere or a single structure. It also outperforms prior work using spectral graph convolution on the same the same tasks, as well as alternative methods that operate on intermediate point cloud representations of 3D shapes. (2) Additionally, visual interpretations for regions on the surface of brain structures that are associated to true positive AD predictions are generated and fall in accordance with the current reports on the structural localization of pathological changes associated to AD. (3) A conditional generative network is also implemented to analyze the effects of phenotypic priors given to the model (i.e. AD diagnosis) in generating subcortical structures. The generated surface meshes by our model indicate learned morphological differences in the presence of AD that agrees with the current literature on patterns of atrophy associated to the disease. In particular, our inference results demonstrate an overall reduction in subcortical mesh volume and surface area in the presence of AD, especially in the hippocampus. The low-dimensional shape descriptors obtained by our generative model are also evaluated in our discriminative baseline comparisons versus our discriminative network and the alternative shape-based approaches.},
author = {Azcona, Emanuel A and Besson, Pierre and Wu, Yunan and Kurani, Ajay S and {Kathleen Bandt}, S and Parrish, Todd B and Katsaggelos, Aggelos K},
doi = {https://doi.org/10.1101/2021.04.15.440008},
journal = {bioRxiv},
keywords = {geometric deep learning,graph convolutional networks,neuroscience,shape analysis},
pages = {2004--2021 , publisher = Cold Spring Harbor Laborat},
title = {{Analyzing Brain Morphology in Alzheimer's Disease Using Discriminative and Generative Spiral Networks}},
url = {https://doi.org/10.1101/2021.04.15.440008},
year = {2021}
}

Several patterns of atrophy have been identified and strongly related to Alzheimer's disease (AD) pathology and its progression. Morphological changes in brain shape have been identified up to ten years before clinical diagnoses of AD, making its early detection more relevant. We propose novel geometric deep learning frameworks for the analysis of brain shape in the context of neurodegeneration caused by AD. Our deep neural networks learn low-dimensional shape descriptors of multiple neuroanatomical structures, instead of handcrafted features for each structure. A discriminative network using spiral convolution on 3D meshes is constructed for the in-vivo binary classification of AD from healthy controls (HCs) using a fast and efficient "spiral" convolution operator on 3D triangular mesh surfaces of human brain subcortical structures extracted from T1-weighted magnetic resonance imaging (MRI). Our network architecture consists of modular learning blocks using residual connections to improve overall classi-fier performance. In this work: (1) a discriminative network is used to analyze the efficacy of disease classification using input data from multiple brain structures and compared to using a single hemisphere or a single structure. It also outperforms prior work using spectral graph convolution on the same the same tasks, as well as alternative methods that operate on intermediate point cloud representations of 3D shapes. (2) Additionally, visual interpretations for regions on the surface of brain structures that are associated to true positive AD predictions are generated and fall in accordance with the current reports on the structural localization of pathological changes associated to AD. (3) A conditional generative network is also implemented to analyze the effects of phenotypic priors given to the model (i.e. AD diagnosis) in generating subcortical structures. The generated surface meshes by our model indicate learned morphological differences in the presence of AD that agrees with the current literature on patterns of atrophy associated to the disease. In particular, our inference results demonstrate an overall reduction in subcortical mesh volume and surface area in the presence of AD, especially in the hippocampus. The low-dimensional shape descriptors obtained by our generative model are also evaluated in our discriminative baseline comparisons versus our discriminative network and the alternative shape-based approaches.

@article{Xin2021,
abstract = {Capturing high-dimensional (HD) data is a long-term challenge in signal processing and related fields. Snapshot compressive imaging (SCI) uses a 2D detector to capture HD (≥3D) data in a snapshot measurement. Via novel optical designs, the 2D detector samples the HD data in a compressive manner; following this, algorithms are employed to reconstruct the desired HD data cube. SCI has been used in hyperspectral imaging, video, holography, tomography, focal depth imaging, polarization imaging, microscopy, and so on. Although the hardware has been investigated for more than a decade, the theoretical guarantees have only recently been derived. Inspired by deep learning, various deep neural networks have also been developed to reconstruct the HD data cube in spectral SCI and video SCI. This article reviews recent advances in SCI hardware, theory, and algorithms, including both optimizationbased and deep learning-based algorithms. Diverse applications and the outlook for SCI are also discussed.},
author = {Yuan, Xin and Brady, David J. and Katsaggelos, Aggelos K.},
doi = {10.1109/MSP.2020.3023869},
issn = {1053-5888},
journal = {IEEE Signal Processing Magazine},
month = {mar},
number = {2},
pages = {65--88},
title = {{Snapshot Compressive Imaging: Theory, Algorithms, and Applications}},
url = {https://ieeexplore.ieee.org/document/9363502/},
volume = {38},
year = {2021}
}

Capturing high-dimensional (HD) data is a long-term challenge in signal processing and related fields. Snapshot compressive imaging (SCI) uses a 2D detector to capture HD (≥3D) data in a snapshot measurement. Via novel optical designs, the 2D detector samples the HD data in a compressive manner; following this, algorithms are employed to reconstruct the desired HD data cube. SCI has been used in hyperspectral imaging, video, holography, tomography, focal depth imaging, polarization imaging, microscopy, and so on. Although the hardware has been investigated for more than a decade, the theoretical guarantees have only recently been derived. Inspired by deep learning, various deep neural networks have also been developed to reconstruct the HD data cube in spectral SCI and video SCI. This article reviews recent advances in SCI hardware, theory, and algorithms, including both optimizationbased and deep learning-based algorithms. Diverse applications and the outlook for SCI are also discussed.

@article{Santiago2021a,
abstract = {Wetlands serve many important ecosystem services, yet the United States lacks up-to-date, high-resolution wetland inventories. New, automated techniques for developing wetland segmentation maps from high-resolution aerial imagery can improve our understanding of the location and amount of wetlands. We assembled training and testing data sets (patch sizes of 28 × 28 m2 and 56 × 56 m2) of high-resolution aerial imagery of wetlands using Illinois Natural History Survey wetland location data and National Agricultural Imagery Project data. Each patch was labeled as wetland or non-wetland. To augment these data sets with additional information, we incorporated digital surface and digital terrain models and topographic wetness index data in the same two patch sizes. Subsequently, we evaluated convolutional neural network (CNN) and Gaussian process-based machine learning methods to produce wetland segmentation maps. We developed the best performing method into a new CNN algorithm, WetSegNet. It exhibited an area under the curve of 98% when used with 56 × 56 m2 patch sizes. WetSegNet developed reliable wetland segmentation maps in test cases in which wetlands would have gone undetected using only the National Land Cover Database. The development of WetSegNet exemplifies the types of data sets and methods that are needed to accelerate the use of high-resolution aerial imagery towards an improved understanding of wetlands. This algorithm could be used by state and federal agencies or other groups to identify wetlands with higher accuracy and at a finer scale than previously possible.},
author = {L{\'{o}}pez-Tapia, Santiago and Ruiz, Pablo and Smith, Mitchell and Matthews, Jeffrey and Zercher, Bradley and Sydorenko, Liliana and Varia, Neelanshi and Jin, Yuanzhe and Wang, Minzi and Dunn, Jennifer B. and Katsaggelos, Aggelos K.},
doi = {10.1016/j.jag.2021.102581},
issn = {15698432},
journal = {International Journal of Applied Earth Observation and Geoinformation},
keywords = {High-resolution aerial imagery,Machine learning,Neural networks,Segmentation,Wetland},
month = {dec},
pages = {102581},
title = {{Machine learning with high-resolution aerial imagery and data fusion to improve and automate the detection of wetlands}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0303243421002889},
volume = {105},
year = {2021}
}

Wetlands serve many important ecosystem services, yet the United States lacks up-to-date, high-resolution wetland inventories. New, automated techniques for developing wetland segmentation maps from high-resolution aerial imagery can improve our understanding of the location and amount of wetlands. We assembled training and testing data sets (patch sizes of 28 × 28 m2 and 56 × 56 m2) of high-resolution aerial imagery of wetlands using Illinois Natural History Survey wetland location data and National Agricultural Imagery Project data. Each patch was labeled as wetland or non-wetland. To augment these data sets with additional information, we incorporated digital surface and digital terrain models and topographic wetness index data in the same two patch sizes. Subsequently, we evaluated convolutional neural network (CNN) and Gaussian process-based machine learning methods to produce wetland segmentation maps. We developed the best performing method into a new CNN algorithm, WetSegNet. It exhibited an area under the curve of 98% when used with 56 × 56 m2 patch sizes. WetSegNet developed reliable wetland segmentation maps in test cases in which wetlands would have gone undetected using only the National Land Cover Database. The development of WetSegNet exemplifies the types of data sets and methods that are needed to accelerate the use of high-resolution aerial imagery towards an improved understanding of wetlands. This algorithm could be used by state and federal agencies or other groups to identify wetlands with higher accuracy and at a finer scale than previously possible.

@article{Emeline2020,
abstract = {A new portable macro X-ray fluorescence scanner has been specifically designed for in situ, real-time elemental mapping of large painted surfaces. This system allows scanning 80 × 80 × 20 cm3 along the X, Z, and Y directions, respectively, with adaptive beam size at the energy of the Rh Ka-line. The detection system consists of a 50 mm2 active area detector coupled to a CUBE pre-amplifier and to the DANTE digital pulse processor (DPP) with adaptive shaping time. The system is controlled with a custom software including a graphical user interface (GUI) programmed in Python for real-time control of the stage, DPP, and camera of the scanner. This system allows considering new ways of sampling the object surface than the usual raster scanning in serpentine as well as a live elaboration of X-ray data; technical details and performances of the scanner are presented in this paper together with an example of its application to investigate painted surface, illustrating the value of the developed instrument.},
author = {Pouyet, Emeline and Barbi, Nicholas and Chopp, Henry and Healy, Owen and Katsaggelos, Aggelos and Moak, Sophia and Mott, Rick and Vermeulen, Marc and Walton, Marc},
doi = {10.1002/xrs.3173},
issn = {0049-8246},
journal = {X-Ray Spectrometry},
month = {aug},
number = {4},
pages = {263--271},
title = {{Development of a highly mobile and versatile large MA‐XRF scanner for in situ analyses of painted work of arts}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/xrs.3173},
volume = {50},
year = {2021}
}

A new portable macro X-ray fluorescence scanner has been specifically designed for in situ, real-time elemental mapping of large painted surfaces. This system allows scanning 80 × 80 × 20 cm3 along the X, Z, and Y directions, respectively, with adaptive beam size at the energy of the Rh Ka-line. The detection system consists of a 50 mm2 active area detector coupled to a CUBE pre-amplifier and to the DANTE digital pulse processor (DPP) with adaptive shaping time. The system is controlled with a custom software including a graphical user interface (GUI) programmed in Python for real-time control of the stage, DPP, and camera of the scanner. This system allows considering new ways of sampling the object surface than the usual raster scanning in serpentine as well as a live elaboration of X-ray data; technical details and performances of the scanner are presented in this paper together with an example of its application to investigate painted surface, illustrating the value of the developed instrument.

@article{Sidi2021,
abstract = {We apply reinforcement learning to video compressive sensing to adapt the compression ratio. Specifically, video snapshot compressive imaging (SCI), which captures high-speed video using a low-speed camera is considered in this work, in which multiple (B) video frames can be reconstructed from a snapshot measurement. One research gap in previous studies is how to adapt B in the video SCI system for different scenes. In this paper, we fill this gap utilizing reinforcement learning (RL). An RL model, as well as various convolutional neural networks for reconstruction, are learned to achieve adaptive sensing of video SCI systems. Furthermore, the performance of an object detection network using directly the video SCI measurements without reconstruction is also used to perform RL-based adaptive video compressive sensing. Our proposed adaptive SCI method can thus be implemented in low cost and real time. Our work takes the technology one step further towards real applications of video SCI.},
archivePrefix = {arXiv},
arxivId = {2105.08205},
author = {Lu, Sidi and Yuan, Xin and Katsaggelos, Aggelos K and Shi, Weisong},
eprint = {2105.08205},
journal = {arXiv preprint arXiv:2105.08205},
month = {may},
title = {{Reinforcement Learning for Adaptive Video Compressive Sensing}},
url = {http://arxiv.org/abs/2105.08205},
year = {2021}
}

We apply reinforcement learning to video compressive sensing to adapt the compression ratio. Specifically, video snapshot compressive imaging (SCI), which captures high-speed video using a low-speed camera is considered in this work, in which multiple (B) video frames can be reconstructed from a snapshot measurement. One research gap in previous studies is how to adapt B in the video SCI system for different scenes. In this paper, we fill this gap utilizing reinforcement learning (RL). An RL model, as well as various convolutional neural networks for reconstruction, are learned to achieve adaptive sensing of video SCI systems. Furthermore, the performance of an object detection network using directly the video SCI measurements without reconstruction is also used to perform RL-based adaptive video compressive sensing. Our proposed adaptive SCI method can thus be implemented in low cost and real time. Our work takes the technology one step further towards real applications of video SCI.

@article{Miguel2021,
abstract = {The volume of labeled data is often the primary determinant of success in developing machine learning algorithms. This has increased interest in methods for leveraging crowds to scale data labeling efforts, and methods to learn from noisy crowd-sourced labels. The need to scale labeling is acute but particularly challenging in medical applications like pathology, due to the expertise required to generate quality labels and the limited availability of qualified experts. In this paper we investigate the application of Scalable Variational Gaussian Processes for Crowdsourcing (SVGPCR) in digital pathology. We compare SVGPCR with other crowdsourcing methods using a large multi-rater dataset where pathologists, pathology residents, and medical students annotated tissue regions breast cancer. Our study shows that SVGPCR is competitive with equivalent methods trained using gold-standard pathologist generated labels, and that SVGPCR meets or exceeds the performance of other crowdsourcing methods based on deep learning. We also show how SVGPCR can effectively learn the class-conditional reliabilities of individual annotators and demonstrate that Gaussian-process classifiers have comparable performance to similar deep learning methods. These results suggest that SVGPCR can meaningfully engage non-experts in pathology labeling tasks, and that the class-conditional reliabilities estimated by SVGPCR may assist in matching annotators to tasks where they perform well.},
author = {L{\'{o}}pez-P{\'{e}}rez, Miguel and Amgad, Mohamed and Morales-{\'{A}}lvarez, Pablo and Ruiz, Pablo and Cooper, Lee A. D. and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1038/s41598-021-90821-3},
issn = {2045-2322},
journal = {Scientific Reports},
month = {jun},
number = {1},
pages = {11612},
pmid = {34078955},
title = {{Learning from crowds in digital pathology using scalable variational Gaussian processes}},
url = {https://www.nature.com/articles/s41598-021-90821-3},
volume = {11},
year = {2021}
}

The volume of labeled data is often the primary determinant of success in developing machine learning algorithms. This has increased interest in methods for leveraging crowds to scale data labeling efforts, and methods to learn from noisy crowd-sourced labels. The need to scale labeling is acute but particularly challenging in medical applications like pathology, due to the expertise required to generate quality labels and the limited availability of qualified experts. In this paper we investigate the application of Scalable Variational Gaussian Processes for Crowdsourcing (SVGPCR) in digital pathology. We compare SVGPCR with other crowdsourcing methods using a large multi-rater dataset where pathologists, pathology residents, and medical students annotated tissue regions breast cancer. Our study shows that SVGPCR is competitive with equivalent methods trained using gold-standard pathologist generated labels, and that SVGPCR meets or exceeds the performance of other crowdsourcing methods based on deep learning. We also show how SVGPCR can effectively learn the class-conditional reliabilities of individual annotators and demonstrate that Gaussian-process classifiers have comparable performance to similar deep learning methods. These results suggest that SVGPCR can meaningfully engage non-experts in pathology labeling tasks, and that the class-conditional reliabilities estimated by SVGPCR may assist in matching annotators to tasks where they perform well.

@article{barutcu2021limited,
abstract = {Computed tomography is a well-established x-ray imaging technique to reconstruct the three-dimensional structure of objects. It has been used extensively in a variety of fields, from diagnostic imaging to materials and biological sciences. One major challenge in some applications, such as in electron or x-ray tomography systems, is that the projections cannot be gathered over all the angles due to the sample holder setup or shape of the sample. This results in an ill-posed problem called the limited angle reconstruction problem. Typical image reconstruction in this setup leads to distortion and artifacts, thereby hindering a quantitative evaluation of the results. To address this challenge, we use a generative model to effectively constrain the solution of a physics-based approach. Our approach is self-training that can iteratively learn the nonlinear mapping from partial projections to the scanned object. Because our approach combines the data likelihood and image prior terms into a single deep network, it is computationally tractable and improves performance through an end-to-end training. We also complement our approach with total-variation regularization to handle high-frequency noise in reconstructions and implement a solver based on alternating direction method of multipliers. We present numerical results for various degrees of missing angle range and noise levels, which demonstrate the effectiveness of the proposed approach.},
author = {Barutcu, Semih and Aslan, Selin and Katsaggelos, Aggelos K. and G{\"{u}}rsoy, Doğa},
doi = {10.1038/s41598-021-97226-2},
issn = {2045-2322},
journal = {Scientific Reports},
month = {sep},
number = {1},
pages = {17740},
pmid = {34489500},
publisher = {Nature Publishing Group UK London},
title = {{Limited-angle computed tomography with deep image and physics priors}},
url = {https://www.nature.com/articles/s41598-021-97226-2},
volume = {11},
year = {2021}
}

Computed tomography is a well-established x-ray imaging technique to reconstruct the three-dimensional structure of objects. It has been used extensively in a variety of fields, from diagnostic imaging to materials and biological sciences. One major challenge in some applications, such as in electron or x-ray tomography systems, is that the projections cannot be gathered over all the angles due to the sample holder setup or shape of the sample. This results in an ill-posed problem called the limited angle reconstruction problem. Typical image reconstruction in this setup leads to distortion and artifacts, thereby hindering a quantitative evaluation of the results. To address this challenge, we use a generative model to effectively constrain the solution of a physics-based approach. Our approach is self-training that can iteratively learn the nonlinear mapping from partial projections to the scanned object. Because our approach combines the data likelihood and image prior terms into a single deep network, it is computationally tractable and improves performance through an end-to-end training. We also complement our approach with total-variation regularization to handle high-frequency noise in reconstructions and implement a solver based on alternating direction method of multipliers. We present numerical results for various degrees of missing angle range and noise levels, which demonstrate the effectiveness of the proposed approach.

@article{KyungPyo2021,
abstract = {Objective. To accelerate compressed sensing (CS) reconstruction of subsampled radial k-space data using a geometrically-derived density compensation function (gDCF) without significant loss in image quality. Approach. We developed a theoretical framework to calculate a gDCF based on Nyquist distance along the radial and circumferential directions of a discrete polar coordinate system. Our gDCF was compared against standard DCF (e.g. ramp filter) and another commonly used DCF (modified Shepp-Logan (SL) filter). The resulting image quality produced by each DCF was quantified using normalized root-mean-square-error (NRMSE), blur metric (1 = blurriest; 0 = sharpest), and structural similarity index (SSIM; 1 = perfect match; 0 = no match) compared with the reference. For filtered backprojection (FBP) of phantom data obtained at the Nyquist sampling rate, Cartesian k-space sampling was used as the reference. For CS reconstruction of subsampled cardiac magnetic resonance imaging datasets (real-time cardiac cine data with 11 projections per frame, n = 20 patients; cardiac perfusion data with 30 projections per frame, n = 19 patients), CS reconstruction without DCF was used as the reference. Main results. The NRMSE, SSIM, and blur metrics of the phantom data were good for all DCFs, confirming that our gDCF produces uniform densities at the upper limit (Nyquist). For CS reconstruction of subsampled real-time cine and cardiac perfusion datasets, the image quality metrics (SSIM, NRMSE) were significantly (p < 0.05) higher for our gDCF than other DCFs, and the reconstruction time was significantly (p < 0.05) faster for our gDCF (reference) than no DCF (11.9%-52.9% slower), standard DCF (23.9%-57.6% slower), and modified SL filter (13.5%-34.8% slower). Significance. The proposed gDCF accelerates CS reconstruction of subsampled radial k-space data without significant loss in image quality compared with no DCF as the reference.},
author = {Hong, KyungPyo and Schiffers, Florian and DiCarlo, Amanda L. and Rigsby, Cynthia K. and Haji-Valizadeh, Hassan and Lee, Daniel C. and Markl, Michael and Katsaggelos, Aggelos K. and Kim, Daniel},
doi = {10.1088/1361-6560/ac2c9d},
issn = {0031-9155},
journal = {Physics in Medicine & Biology},
keywords = {Compressed sensing,Density compensation,Filtered backprojection,Image reconstruction,MRI,Radial k-space sampling},
month = {nov},
number = {21},
pages = {21NT01},
title = {{Accelerating compressed sensing reconstruction of subsampled radial k-space data using geometrically-derived density compensation}},
url = {https://iopscience.iop.org/article/10.1088/1361-6560/ac2c9d},
volume = {66},
year = {2021}
}

Objective. To accelerate compressed sensing (CS) reconstruction of subsampled radial k-space data using a geometrically-derived density compensation function (gDCF) without significant loss in image quality. Approach. We developed a theoretical framework to calculate a gDCF based on Nyquist distance along the radial and circumferential directions of a discrete polar coordinate system. Our gDCF was compared against standard DCF (e.g. ramp filter) and another commonly used DCF (modified Shepp-Logan (SL) filter). The resulting image quality produced by each DCF was quantified using normalized root-mean-square-error (NRMSE), blur metric (1 = blurriest; 0 = sharpest), and structural similarity index (SSIM; 1 = perfect match; 0 = no match) compared with the reference. For filtered backprojection (FBP) of phantom data obtained at the Nyquist sampling rate, Cartesian k-space sampling was used as the reference. For CS reconstruction of subsampled cardiac magnetic resonance imaging datasets (real-time cardiac cine data with 11 projections per frame, n = 20 patients; cardiac perfusion data with 30 projections per frame, n = 19 patients), CS reconstruction without DCF was used as the reference. Main results. The NRMSE, SSIM, and blur metrics of the phantom data were good for all DCFs, confirming that our gDCF produces uniform densities at the upper limit (Nyquist). For CS reconstruction of subsampled real-time cine and cardiac perfusion datasets, the image quality metrics (SSIM, NRMSE) were significantly (p < 0.05) higher for our gDCF than other DCFs, and the reconstruction time was significantly (p < 0.05) faster for our gDCF (reference) than no DCF (11.9%-52.9% slower), standard DCF (23.9%-57.6% slower), and modified SL filter (13.5%-34.8% slower). Significance. The proposed gDCF accelerates CS reconstruction of subsampled radial k-space data without significant loss in image quality compared with no DCF as the reference.

@article{Santiago2021,
abstract = {In recent years, deep learning-based models have gained momentum in imaging problems such as image and video super-resolution, image restoration or inpainting. The analytical approaches that have traditionally been used to solve image inverse problems have started to be replaced by deep learning ones, being outperformed in terms of efficacy and efficiency in many applications. However, deep learning-based models lack the adaptability of analytical models, thus making them unsuitable for dealing simultaneously with different forward image formation models. In contrast to analytical methods, deep learning models typically do not use domain knowledge and rely on learning the solution to the inverse problem from large data sets. This is making them susceptible to errors caused by the presence of degradations not seen during training. Hybrid models combining analytical and deep learning approaches have been introduced to solve such generalization issues while retaining the efficacy of deep learning models. In this work, we review deep learning and hybrid methods for solving imaging inverse problems, focusing on image and video super-resolution and image restoration. Furthermore, we discuss open problems in this area that would be of critical importance in the future, the challenges of applying deep learning models to solve them, and how future research could address them.},
author = {L{\'{o}}pez-Tapia, Santiago and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.dsp.2021.103285},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Convolutional neural network,Deep learning,Inverse imaging problems,Video super-resolution},
month = {dec},
pages = {103285},
title = {{Deep learning approaches to inverse problems in imaging: Past, present and future}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200421003249},
volume = {119},
year = {2021}
}

In recent years, deep learning-based models have gained momentum in imaging problems such as image and video super-resolution, image restoration or inpainting. The analytical approaches that have traditionally been used to solve image inverse problems have started to be replaced by deep learning ones, being outperformed in terms of efficacy and efficiency in many applications. However, deep learning-based models lack the adaptability of analytical models, thus making them unsuitable for dealing simultaneously with different forward image formation models. In contrast to analytical methods, deep learning models typically do not use domain knowledge and rely on learning the solution to the inverse problem from large data sets. This is making them susceptible to errors caused by the presence of degradations not seen during training. Hybrid models combining analytical and deep learning approaches have been introduced to solve such generalization issues while retaining the efficacy of deep learning models. In this work, we review deep learning and hybrid methods for solving imaging inverse problems, focusing on image and video super-resolution and image restoration. Furthermore, we discuss open problems in this area that would be of critical importance in the future, the challenges of applying deep learning models to solve them, and how future research could address them.

@article{Amil2021,
author = {Amil, Dravid and Aggelos, K Katsaggelos},
title = {{Visual explanations for convolutional neural networks via latent traversal}},
year = {2021}
}

@article{Kevin2019,
abstract = {We present the design of a citizen science system that uses machine learning to guide the presentation of image classification tasks to newcomers to help them more quickly learn how to do the task while still contributing to the work of the project. A Bayesian model for tracking volunteer learning for training with tasks with uncertain outcomes is presented and fit to data from 12,986 volunteer contributors. The model can be used both to estimate the ability of volunteers and to decide the classification of an image. A simulation of the model applied to volunteer promotion and image retirement suggests that the model requires fewer classifications than the current system.},
author = {Crowston, Kevin and Osterlund, Carsten and Lee, Tae Kyoung and Jackson, Corey and Harandi, Mahboobeh and Allen, Sarah and Bahaadini, Sara and Coughlin, Scott and Katsaggelos, Aggelos K. and Larson, Shane L. and Rohani, Neda and Smith, Joshua R. and Trouille, Laura and Zevin, Michael},
doi = {10.1109/TLT.2019.2936480},
issn = {1939-1382},
journal = {IEEE Transactions on Learning Technologies},
keywords = {Citizen science,machine learning,training},
month = {jan},
number = {1},
pages = {123--134},
title = {{Knowledge Tracing to Model Learning in Online Citizen Science Projects}},
url = {https://ieeexplore.ieee.org/document/8812979/},
volume = {13},
year = {2020}
}

We present the design of a citizen science system that uses machine learning to guide the presentation of image classification tasks to newcomers to help them more quickly learn how to do the task while still contributing to the work of the project. A Bayesian model for tracking volunteer learning for training with tasks with uncertain outcomes is presented and fit to data from 12,986 volunteer contributors. The model can be used both to estimate the ability of volunteers and to decide the classification of an image. A simulation of the model applied to volunteer promotion and image retirement suggests that the model requires fewer classifications than the current system.

@article{Kuan2020,
abstract = {Light field microscopy (LFM) is an emerging technology for high-speed wide-field 3D imaging by capturing 4D light field of 3D volumes. However, its 3D imaging capability comes at a cost of lateral resolution. In addition, the lateral resolution is not uniform across depth in the light field dconvolution reconstructions. To address these problems, here, we propose a snapshot multifocal light field microscopy (MFLFM) imaging method. The underlying concept of the MFLFM is to collect multiple focal shifted light fields simultaneously. We show that by focal stacking those focal shifted light fields, the depth-of-field (DOF) of the LFM can be further improved but without sacrificing the lateral resolution. Also, if all differently focused light fields are utilized together in the deconvolution, the MFLFM could achieve a high and uniform lateral resolution within a larger DOF. We present a house-built MFLFM system by placing a diffractive optical element at the Fourier plane of a conventional LFM. The optical performance of the MFLFM are analyzed and given. Both simulations and proof-of-principle experimental results are provided to demonstrate the effectiveness and benefits of the MFLFM. We believe that the proposed snapshot MFLFM has potential to enable high-speed and high resolution 3D imaging applications.},
author = {He, Kuan and Wang, Xiaolei and Wang, Zihao W. and Yi, Hannah and Scherer, Norbert F. and Katsaggelos, Aggelos K. and Cossairt, Oliver},
doi = {10.1364/OE.390719},
issn = {1094-4087},
journal = {Optics Express},
month = {apr},
number = {8},
pages = {12108},
pmid = {32403711},
title = {{Snapshot multifocal light field microscopy}},
url = {https://opg.optica.org/abstract.cfm?URI=oe-28-8-12108},
volume = {28},
year = {2020}
}

Light field microscopy (LFM) is an emerging technology for high-speed wide-field 3D imaging by capturing 4D light field of 3D volumes. However, its 3D imaging capability comes at a cost of lateral resolution. In addition, the lateral resolution is not uniform across depth in the light field dconvolution reconstructions. To address these problems, here, we propose a snapshot multifocal light field microscopy (MFLFM) imaging method. The underlying concept of the MFLFM is to collect multiple focal shifted light fields simultaneously. We show that by focal stacking those focal shifted light fields, the depth-of-field (DOF) of the LFM can be further improved but without sacrificing the lateral resolution. Also, if all differently focused light fields are utilized together in the deconvolution, the MFLFM could achieve a high and uniform lateral resolution within a larger DOF. We present a house-built MFLFM system by placing a diffractive optical element at the Fourier plane of a conventional LFM. The optical performance of the MFLFM are analyzed and given. Both simulations and proof-of-principle experimental results are provided to demonstrate the effectiveness and benefits of the MFLFM. We believe that the proposed snapshot MFLFM has potential to enable high-speed and high resolution 3D imaging applications.

@article{Yunan2020,
abstract = {Brain structure is tightly coupled with brain functions, but it remains unclear how cognition is related to brain morphology, and what is consistent across neurodevelopment. In this work, we developed graph convolutional neural networks (gCNNs) to predict Fluid Intelligence (Gf) from shapes of cortical ribbons and subcortical structures. T1-weighted MRIs from two independent cohorts, the Human Connectome Project (HCP; age: 28.81±3.70) and the Adolescent Brain Cognitive Development Study (ABCD; age: 9.93±0.62) were independently analyzed. Cortical and subcortical surfaces were extracted and modeled as surface meshes. Three gCNNs were trained and evaluated using six-fold nested cross-validation. Overall, combining cortical and subcortical surfaces yielded the best predictions on both HCP (R=0.454) and ABCD datasets (R=0.314), and outperformed the current literature. Across both datasets, the morphometry of the amygdala and hippocampus, along with temporal, parietal and cingulate cortex consistently drove the prediction of Gf, suggesting a novel reframing of the morphometry underlying Gf.. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted October 15, 2020. ; https://doi.},
author = {Wu, Yunan and Besson, Pierre and Azcona, Emanuel A and Bandt, S Kathleen and Parrish, Todd B and Breiter, Hans C and Katsaggelos, Aggelos K},
doi = {10.1101/2020.10.14.331199},
journal = {bioRxiv},
pages = {2020.10.14.331199},
title = {{Novel age-dependent cortico-subcortical morphologic interactions predict fluid intelligence: A multi-cohort geometric deep learning study Authors}},
url = {https://doi.org/10.1101/2020.10.14.331199},
year = {2020}
}

Brain structure is tightly coupled with brain functions, but it remains unclear how cognition is related to brain morphology, and what is consistent across neurodevelopment. In this work, we developed graph convolutional neural networks (gCNNs) to predict Fluid Intelligence (Gf) from shapes of cortical ribbons and subcortical structures. T1-weighted MRIs from two independent cohorts, the Human Connectome Project (HCP; age: 28.81±3.70) and the Adolescent Brain Cognitive Development Study (ABCD; age: 9.93±0.62) were independently analyzed. Cortical and subcortical surfaces were extracted and modeled as surface meshes. Three gCNNs were trained and evaluated using six-fold nested cross-validation. Overall, combining cortical and subcortical surfaces yielded the best predictions on both HCP (R=0.454) and ABCD datasets (R=0.314), and outperformed the current literature. Across both datasets, the morphometry of the amygdala and hippocampus, along with temporal, parietal and cingulate cortex consistently drove the prediction of Gf, suggesting a novel reframing of the morphometry underlying Gf.. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted October 15, 2020. ; https://doi.

@article{Alexis2020,
abstract = {3D shape reconstruction is a primary component of augmented/virtual reality. Despite being highly advanced, existing solutions based on RGB, RGB-D and Lidar sensors are power and data intensive, which introduces challenges for deployment in edge devices. We approach 3D reconstruction with an event camera, a sensor with significantly lower power, latency and data expense while enabling high dynamic range. While previous event-based 3D reconstruction methods are primarily based on stereo vision, we cast the problem as multi-view shape from silhouette using a monocular event camera. The output from a moving event camera is a sparse point set of space-time gradients, largely sketching scene/object edges and contours. We first introduce an event-to-silhouette (E2S) neural network module to transform a stack of event frames to the corresponding silhouettes, with additional neural branches for camera pose regression. Second, we introduce E3D, which employs a 3D differentiable renderer (PyTorch3D) to enforce cross-view 3D mesh consistency and fine-tune the E2S and pose network. Lastly, we introduce a 3D-to-events simulation pipeline and apply it to publicly available object datasets and generate synthetic event/silhouette training pairs for supervised learning.},
archivePrefix = {arXiv},
arxivId = {2012.05214},
author = {Baudron, Alexis and Wang, Zihao W. and Cossairt, Oliver and Katsaggelos, Aggelos K.},
eprint = {2012.05214},
journal = {arXiv preprint arXiv:2012.05214},
month = {dec},
title = {{E3D: Event-Based 3D Shape Reconstruction}},
url = {http://arxiv.org/abs/2012.05214},
year = {2020}
}

3D shape reconstruction is a primary component of augmented/virtual reality. Despite being highly advanced, existing solutions based on RGB, RGB-D and Lidar sensors are power and data intensive, which introduces challenges for deployment in edge devices. We approach 3D reconstruction with an event camera, a sensor with significantly lower power, latency and data expense while enabling high dynamic range. While previous event-based 3D reconstruction methods are primarily based on stereo vision, we cast the problem as multi-view shape from silhouette using a monocular event camera. The output from a moving event camera is a sparse point set of space-time gradients, largely sketching scene/object edges and contours. We first introduce an event-to-silhouette (E2S) neural network module to transform a stack of event frames to the corresponding silhouettes, with additional neural branches for camera pose regression. Second, we introduce E3D, which employs a 3D differentiable renderer (PyTorch3D) to enforce cross-view 3D mesh consistency and fine-tune the E2S and pose network. Lastly, we introduce a 3D-to-events simulation pipeline and apply it to publicly available object datasets and generate synthetic event/silhouette training pairs for supervised learning.

@article{Florian2020,
abstract = {We introduce a system that exploits the screen and front-facing camera of a mobile device to perform three-dimensional deflectometry-based surface measurements. In contrast to current mobile deflectometry systems, our method can capture surfaces with large normal variation and wide field of view (FoV). We achieve this by applying automated multi-view panoramic stitching algorithms to produce a large FoV normal map from a hand-guided capture process without the need for external tracking systems, like robot arms or fiducials. The presented work enables 3D surface measurements of specular objects 'in the wild' with a system accessible to users with little to no technical imaging experience. We demonstrate high-quality 3D surface measurements without the need for a calibration procedure. We provide experimental results with our prototype Deflectometry system and discuss applications for computer vision tasks such as object detection and recognition.},
author = {Willomitzer, Florian and Yeh, Chia-Kai and Gupta, Vikas and Spies, William and Schiffers, Florian and Katsaggelos, Aggelos and Walton, Marc and Cossairt, Oliver},
doi = {10.1364/OE.383475},
issn = {1094-4087},
journal = {Optics Express},
month = {mar},
number = {7},
pages = {9027},
pmid = {32225516},
title = {{Hand-guided qualitative deflectometry with a mobile device}},
url = {https://opg.optica.org/abstract.cfm?URI=oe-28-7-9027},
volume = {28},
year = {2020}
}

We introduce a system that exploits the screen and front-facing camera of a mobile device to perform three-dimensional deflectometry-based surface measurements. In contrast to current mobile deflectometry systems, our method can capture surfaces with large normal variation and wide field of view (FoV). We achieve this by applying automated multi-view panoramic stitching algorithms to produce a large FoV normal map from a hand-guided capture process without the need for external tracking systems, like robot arms or fiducials. The presented work enables 3D surface measurements of specular objects 'in the wild' with a system accessible to users with little to no technical imaging experience. We demonstrate high-quality 3D surface measurements without the need for a calibration procedure. We provide experimental results with our prototype Deflectometry system and discuss applications for computer vision tasks such as object detection and recognition.

@article{Srutarshi2020,
abstract = {With several advantages over conventional RGB cameras, event cameras have provided new opportunities for tackling visual tasks under challenging scenarios with fast motion, high dynamic range, and/or power constraint. Yet unlike image/video compression, the performance of event compression algorithm is far from satisfying and practical. The main challenge for compressing events is the unique event data form, i.e., a stream of asynchronously fired event tuples each encoding the 2D spatial location, timestamp, and polarity (denoting an increase or decrease in brightness). Since events only encode temporal variations, they lack spatial structure which is crucial for compression. To address this problem, we propose a novel event compression algorithm based on a quad tree (QT) segmentation map derived from the adjacent intensity images. The QT informs 2D spatial priority within the 3D space-time volume. In the event encoding step, events are first aggregated over time to form polarity-based event histograms. The histograms are then variably sampled via Poisson Disk Sampling prioritized by the QT based segmentation map. Next, differential encoding and run length encoding are employed for encoding the spatial and polarity information of the sampled events, respectively, followed by Huffman encoding to produce the final encoded events. Our Poisson Disk Sampling based Lossy Event Compression (PDS-LEC) algorithm performs rate-distortion based optimal allocation. On average, our algorithm achieves greater than 6x compression compared to the state of the art.},
archivePrefix = {arXiv},
arxivId = {2005.00974},
author = {Banerjee, Srutarshi and Wang, Zihao W. and Chopp, Henry H. and Cossairt, Oliver and Katsaggelos, Aggelos},
eprint = {2005.00974},
journal = {arXiv},
keywords = {Event cameras,Event-based vision,Image and video compression},
month = {may},
title = {{Lossy Event Compression based on Image-derived Quad Trees and Poisson Disk Sampling}},
url = {http://arxiv.org/abs/2005.00974},
year = {2020}
}

With several advantages over conventional RGB cameras, event cameras have provided new opportunities for tackling visual tasks under challenging scenarios with fast motion, high dynamic range, and/or power constraint. Yet unlike image/video compression, the performance of event compression algorithm is far from satisfying and practical. The main challenge for compressing events is the unique event data form, i.e., a stream of asynchronously fired event tuples each encoding the 2D spatial location, timestamp, and polarity (denoting an increase or decrease in brightness). Since events only encode temporal variations, they lack spatial structure which is crucial for compression. To address this problem, we propose a novel event compression algorithm based on a quad tree (QT) segmentation map derived from the adjacent intensity images. The QT informs 2D spatial priority within the 3D space-time volume. In the event encoding step, events are first aggregated over time to form polarity-based event histograms. The histograms are then variably sampled via Poisson Disk Sampling prioritized by the QT based segmentation map. Next, differential encoding and run length encoding are employed for encoding the spatial and polarity information of the sampled events, respectively, followed by Huffman encoding to produce the final encoded events. Our Poisson Disk Sampling based Lossy Event Compression (PDS-LEC) algorithm performs rate-distortion based optimal allocation. On average, our algorithm achieves greater than 6x compression compared to the state of the art.

@article{Florian2020a,
abstract = {Imaging through scattering media finds applications in diverse fields from biomedicine to autonomous driving. However, interpreting the resulting images is difficult due to blur caused by the scattering of photons within the medium. Transient information, captured with fast temporal sensors, can be used to significantly improve the quality of images acquired in scattering conditions. Photon scattering, within a highly scattering media, is well modeled by the diffusion approximation of the Radiative Transport Equation (RTE). Its solution is easily derived which can be interpreted as a Spatio-Temporal Point Spread Function (ST-PSF). In this paper, we first discuss the properties of the ST-PSF and subsequently use this knowledge to simulate transient imaging through highly scattering media. We then propose a framework to invert the forward model, which assumes Poisson noise, to recover a noise-free, unblurred image by solving an optimization problem.},
author = {Schiffers, Florian and Fiske, Lionel and Ruiz, Pablo and Katsaggelos, Aggelos K. and Cossairt, Oliver},
doi = {10.2352/ISSN.2470-1173.2020.14.COIMG-306},
issn = {2470-1173},
journal = {Electronic Imaging},
month = {jan},
number = {14},
pages = {306--1--306--6},
title = {{Imaging through Scattering Media with a Learning Based Prior}},
url = {https://library.imaging.org/ei/articles/32/14/art00012},
volume = {32},
year = {2020}
}

Imaging through scattering media finds applications in diverse fields from biomedicine to autonomous driving. However, interpreting the resulting images is difficult due to blur caused by the scattering of photons within the medium. Transient information, captured with fast temporal sensors, can be used to significantly improve the quality of images acquired in scattering conditions. Photon scattering, within a highly scattering media, is well modeled by the diffusion approximation of the Radiative Transport Equation (RTE). Its solution is easily derived which can be interpreted as a Spatio-Temporal Point Spread Function (ST-PSF). In this paper, we first discuss the properties of the ST-PSF and subsequently use this knowledge to simulate transient imaging through highly scattering media. We then propose a framework to invert the forward model, which assumes Poisson noise, to recover a noise-free, unblurred image by solving an optimization problem.

@article{Perez-Bueno2020,
abstract = {In digital histopathological image analysis, two conflicting objectives are often pursued: closeness to the original tissue and high classification performance. The former objective tries to recover images (stains) that are as close as possible to the ones obtained by staining the tissue with a single dye. The latter objective requires images that allow the extraction of better features for an improved classification, even if their appearance is not close to single stained tissues. In this paper we propose a framework that achieves both objectives depending on the number of stains used to mathematically decompose the scanned image. The proposed framework uses a total variation prior for each stain together with the similarity to a given reference color-vector matrix. Variational inference and an evidence lower bound are utilized to automatically estimate all the latent variables and model parameters. The proposed methodology is tested on real images and compared to classical and state-of-the-art methods for histopathological blind image color deconvolution and prostate cancer classification.},
author = {P{\'{e}}rez-Bueno, Fernando and L{\'{o}}pez-P{\'{e}}rez, Miguel and Vega, Miguel and Mateos, Javier and Naranjo, Valery and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.dsp.2020.102727},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Blind color deconvolution,Histopathological images,Prostate cancer,Variational Bayes},
month = {jun},
pages = {102727},
title = {{A TV-based image processing framework for blind color deconvolution and classification of histological images}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200420300725},
volume = {101},
year = {2020}
}

In digital histopathological image analysis, two conflicting objectives are often pursued: closeness to the original tissue and high classification performance. The former objective tries to recover images (stains) that are as close as possible to the ones obtained by staining the tissue with a single dye. The latter objective requires images that allow the extraction of better features for an improved classification, even if their appearance is not close to single stained tissues. In this paper we propose a framework that achieves both objectives depending on the number of stains used to mathematically decompose the scanned image. The proposed framework uses a total variation prior for each stain together with the similarity to a given reference color-vector matrix. Variational inference and an evidence lower bound are utilized to automatically estimate all the latent variables and model parameters. The proposed methodology is tested on real images and compared to classical and state-of-the-art methods for histopathological blind image color deconvolution and prostate cancer classification.

@article{Hassan2020,
abstract = {Purpose: To implement an integrated reconstruction pipeline including a graphics processing unit (GPU)–based convolutional neural network (CNN) architecture and test whether it reconstructs four-dimensional non-Cartesian, non–contrast material–enhanced MR angiographic k-space data faster than a central processing unit (CPU)–based compressed sensing (CS) reconstruction pipeline, without significant losses in data fidelity, summed visual score (SVS), or arterial vessel–diameter measurements. Materials and Methods: Raw k-space data of 24 patients (18 men and six women; mean age, 56.8 years 6 11.8 [standard deviation]) suspected of having thoracic aortic disease were used to evaluate the proposed reconstruction pipeline derived from an open-source threedimensional CNN. For training, 4800 zero-filled images and the corresponding CS-reconstructed images from 10 patients were used as input-output pairs. For testing, 6720 zero-filled images from 14 different patients were used as inputs to a trained CNN. Metrics for evaluating the agreement between the CNN and CS images included reconstruction times, structural similarity index (SSIM) and normalized root-mean-square error (NRMSE), SVS (3 = nondiagnostic, 9 = clinically acceptable, 15 = excellent), and vessel diameters. Results: The mean reconstruction time was 65 times and 69 times shorter for the CPU-based and GPU-based CNN pipelines (216.6 seconds 6 40.5 and 204.9 seconds 6 40.5), respectively, than for CS (14 152.3 seconds 6 1708.6) (P, .001). Compared with CS as practical ground truth, CNNs produced high data fidelity (SSIM = 0.94 6 0.02, NRMSE = 2.8% 6 0.4) and not significantly different (P = .25) SVS and aortic diameters, except at one out of seven locations, where the percentage difference was only 3% (ie, clinically irrelevant). Conclusion: The proposed integrated reconstruction pipeline including a CNN architecture is capable of rapidly reconstructing timeresolved volumetric cardiovascular MRI k-space data, without a significant loss in data quality, thereby supporting clinical translation of said non–contrast-enhanced MR angiograms.},
author = {Haji-Valizadeh, Hassan and Shen, Daming and Avery, Ryan J. and Serhal, Ali M. and Schiffers, Florian A. and Katsaggelos, Aggelos K. and Cossairt, Oliver S. and Kim, Daniel},
doi = {10.1148/ryct.2020190205},
issn = {2638-6135},
journal = {Radiology: Cardiothoracic Imaging},
month = {jun},
number = {3},
pages = {e190205},
title = {{Rapid Reconstruction of Four-dimensional MR Angiography of the Thoracic Aorta Using a Convolutional Neural Network}},
url = {http://pubs.rsna.org/doi/10.1148/ryct.2020190205},
volume = {2},
year = {2020}
}

Purpose: To implement an integrated reconstruction pipeline including a graphics processing unit (GPU)–based convolutional neural network (CNN) architecture and test whether it reconstructs four-dimensional non-Cartesian, non–contrast material–enhanced MR angiographic k-space data faster than a central processing unit (CPU)–based compressed sensing (CS) reconstruction pipeline, without significant losses in data fidelity, summed visual score (SVS), or arterial vessel–diameter measurements. Materials and Methods: Raw k-space data of 24 patients (18 men and six women; mean age, 56.8 years 6 11.8 [standard deviation]) suspected of having thoracic aortic disease were used to evaluate the proposed reconstruction pipeline derived from an open-source threedimensional CNN. For training, 4800 zero-filled images and the corresponding CS-reconstructed images from 10 patients were used as input-output pairs. For testing, 6720 zero-filled images from 14 different patients were used as inputs to a trained CNN. Metrics for evaluating the agreement between the CNN and CS images included reconstruction times, structural similarity index (SSIM) and normalized root-mean-square error (NRMSE), SVS (3 = nondiagnostic, 9 = clinically acceptable, 15 = excellent), and vessel diameters. Results: The mean reconstruction time was 65 times and 69 times shorter for the CPU-based and GPU-based CNN pipelines (216.6 seconds 6 40.5 and 204.9 seconds 6 40.5), respectively, than for CS (14 152.3 seconds 6 1708.6) (P, .001). Compared with CS as practical ground truth, CNNs produced high data fidelity (SSIM = 0.94 6 0.02, NRMSE = 2.8% 6 0.4) and not significantly different (P = .25) SVS and aortic diameters, except at one out of seven locations, where the percentage difference was only 3% (ie, clinically irrelevant). Conclusion: The proposed integrated reconstruction pipeline including a CNN architecture is capable of rapidly reconstructing timeresolved volumetric cardiovascular MRI k-space data, without a significant loss in data quality, thereby supporting clinical translation of said non–contrast-enhanced MR angiograms.

@article{Arun2020,
abstract = {In this paper, we present a temporal capsule network architecture to encode motion in videos as an instantiation parameter. The extracted motion is used to perform motion-compensated error concealment. We modify the original architecture and use a carefully curated dataset to enable the training of capsules spatially and temporally. First, we add the temporal dimension by taking co-located “patches” from three consecutive frames obtained from standard video sequences to form input data “cubes.” Second, the network is designed with an initial feature extraction layer that operates on all three dimensions to generate spatiotemporal features. Additionally, we implement the PrimaryCaps module with a recurrent layer, instead of a conventional convolutional layer, to extract short-term motion-related temporal dependencies and encode them as activation vectors in the capsule output. Finally, the capsule output is combined with the most-recent past frame and passed through a fully connected reconstruction network to perform motion-compensated error concealment. We study the effectiveness of temporal capsules by comparing the proposed model with architectures that do not include capsules. Although the quality of the reconstruction shows room for improvement, we successfully demonstrate that capsules-based architectures can be designed to operate in the temporal dimension to encode motion-related attributes as instantiation parameters. The accuracy of motion estimation is evaluated by comparing both the reconstructed frame outputs and the corresponding optical flow estimates with ground truth data.},
author = {Sankisa, Arun and Punjabi, Arjun and Katsaggelos, Aggelos K.},
doi = {10.1007/s11760-020-01671-x},
issn = {1863-1703},
journal = {Signal, Image and Video Processing},
keywords = {Capsule networks,Conv3D,ConvLSTM,Error concealment,Motion estimation},
month = {oct},
number = {7},
pages = {1369--1377},
title = {{Temporal capsule networks for video motion estimation and error concealment}},
url = {http://link.springer.com/10.1007/s11760-020-01671-x},
volume = {14},
year = {2020}
}

In this paper, we present a temporal capsule network architecture to encode motion in videos as an instantiation parameter. The extracted motion is used to perform motion-compensated error concealment. We modify the original architecture and use a carefully curated dataset to enable the training of capsules spatially and temporally. First, we add the temporal dimension by taking co-located “patches” from three consecutive frames obtained from standard video sequences to form input data “cubes.” Second, the network is designed with an initial feature extraction layer that operates on all three dimensions to generate spatiotemporal features. Additionally, we implement the PrimaryCaps module with a recurrent layer, instead of a conventional convolutional layer, to extract short-term motion-related temporal dependencies and encode them as activation vectors in the capsule output. Finally, the capsule output is combined with the most-recent past frame and passed through a fully connected reconstruction network to perform motion-compensated error concealment. We study the effectiveness of temporal capsules by comparing the proposed model with architectures that do not include capsules. Although the quality of the reconstruction shows room for improvement, we successfully demonstrate that capsules-based architectures can be designed to operate in the temporal dimension to encode motion-related attributes as instantiation parameters. The accuracy of motion estimation is evaluated by comparing both the reconstructed frame outputs and the corresponding optical flow estimates with ground truth data.

@article{Natalia2019,
abstract = {Most whole-slide histological images are stained with two or more chemical dyes. Slide stain separation or color deconvolution is a crucial step within the digital pathology workflow. In this paper, the blind color deconvolution problem is formulated within the Bayesian framework. Starting from a multi-stained histological image, our model takes into account both spatial relations among the concentration image pixels and similarity between a given reference color-vector matrix and the estimated one. Using Variational Bayes inference, three efficient new blind color deconvolution methods are proposed which provide automated procedures to estimate all the model parameters in the problem. A comparison with classical and current state-of-the-art color deconvolution algorithms using real images has been carried out demonstrating the superiority of the proposed approach.},
author = {Hidalgo-Gavira, Natalia and Mateos, Javier and Vega, Miguel and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2019.2946442},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian modeling and inference,Blind color deconvolution,histopathological images,variational Bayes},
number = {1},
pages = {2026--2036},
pmid = {31634128},
title = {{Variational Bayesian Blind Color Deconvolution of Histopathological Images}},
url = {https://ieeexplore.ieee.org/document/8870230/},
volume = {29},
year = {2020}
}

Most whole-slide histological images are stained with two or more chemical dyes. Slide stain separation or color deconvolution is a crucial step within the digital pathology workflow. In this paper, the blind color deconvolution problem is formulated within the Bayesian framework. Starting from a multi-stained histological image, our model takes into account both spatial relations among the concentration image pixels and similarity between a given reference color-vector matrix and the estimated one. Using Variational Bayes inference, three efficient new blind color deconvolution methods are proposed which provide automated procedures to estimate all the model parameters in the problem. A comparison with classical and current state-of-the-art color deconvolution algorithms using real images has been carried out demonstrating the superiority of the proposed approach.

@article{Fernando2020b,
abstract = {Pansharpening is a technique that fuses a low spatial resolution multispectral image and a high spatial resolution panchromatic one to obtain a multispectral image with the spatial resolution of the latter while preserving the spectral information of the multispectral image. In this paper we propose a variational Bayesian methodology for pansharpening. The proposed methodology uses the sensor characteristics to model the observation process and Super-Gaussian sparse image priors on the expected characteristics of the pansharpened image. The pansharpened image, as well as all model and variational parameters, are estimated within the proposed methodology. Using real and synthetic data, the quality of the pansharpened images is assessed both visually and quantitatively and compared with other pansharpening methods. Theoretical and experimental results demonstrate the effectiveness, efficiency, and flexibility of the proposed formulation.},
author = {P{\'{e}}rez-Bueno, Fernando and Vega, Miguel and Mateos, Javier and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.3390/s20185308},
issn = {1424-8220},
journal = {Sensors},
keywords = {Image fusion,Pansharpening,Super-Gaussians,Variational Bayesian},
month = {sep},
number = {18},
pages = {5308},
pmid = {32948056},
title = {{Variational Bayesian Pansharpening with Super-Gaussian Sparse Image Priors}},
url = {https://www.mdpi.com/1424-8220/20/18/5308},
volume = {20},
year = {2020}
}

Pansharpening is a technique that fuses a low spatial resolution multispectral image and a high spatial resolution panchromatic one to obtain a multispectral image with the spatial resolution of the latter while preserving the spectral information of the multispectral image. In this paper we propose a variational Bayesian methodology for pansharpening. The proposed methodology uses the sensor characteristics to model the observation process and Super-Gaussian sparse image priors on the expected characteristics of the pansharpened image. The pansharpened image, as well as all model and variational parameters, are estimated within the proposed methodology. Using real and synthetic data, the quality of the pansharpened images is assessed both visually and quantitatively and compared with other pansharpening methods. Theoretical and experimental results demonstrate the effectiveness, efficiency, and flexibility of the proposed formulation.

@article{Arjun2020,
abstract = {Capsule networks are a recently developed class of neural networks that potentially address some of the deficiencies with traditional convolutional neural networks. By replacing the standard scalar activations with vectors, and by connecting the artificial neurons in a new way, capsule networks aim to be the next great development for computer vision applications. However, in order to determine whether these networks truly operate differently than traditional networks, one must look at the differences in the capsule features. To this end, we perform several analyses with the purpose of elucidating capsule features and determining whether they perform as described in the initial publication. First, we perform a deep visualization analysis to visually compare capsule features and convolutional neural network features. Then, we look at the ability for capsule features to encode information across the vector components and address what changes in the capsule architecture provides the most benefit. Finally, we look at how well the capsule features are able to encode instantiation parameters of class objects via visual transformations.},
archivePrefix = {arXiv},
arxivId = {2001.10964},
author = {Punjabi, Arjun and Schmid, Jonas and Katsaggelos, Aggelos K.},
eprint = {2001.10964},
journal = {arXiv preprint arXiv:2001.10964},
month = {jan},
title = {{Examining the Benefits of Capsule Neural Networks}},
url = {http://arxiv.org/abs/2001.10964},
year = {2020}
}

Capsule networks are a recently developed class of neural networks that potentially address some of the deficiencies with traditional convolutional neural networks. By replacing the standard scalar activations with vectors, and by connecting the artificial neurons in a new way, capsule networks aim to be the next great development for computer vision applications. However, in order to determine whether these networks truly operate differently than traditional networks, one must look at the differences in the capsule features. To this end, we perform several analyses with the purpose of elucidating capsule features and determining whether they perform as described in the initial publication. First, we perform a deep visualization analysis to visually compare capsule features and convolutional neural network features. Then, we look at the ability for capsule features to encode information across the vector components and address what changes in the capsule architecture provides the most benefit. Finally, we look at how well the capsule features are able to encode instantiation parameters of class objects via visual transformations.

@article{bae2020relating,
author = {Bae, Jinhyeong and Stocks, Jane and Heywood, Ashley and Jung, Youngmoon and Katsaggelos, Aggelos and Jenkins, Lisanne M and Karteek, Popuri and Beg, Mirza Faisal and Wang, Lei},
doi = {10.1002/alz.043538},
institution = {ALZ},
issn = {1552-5260},
journal = {Alzheimer's & Dementia},
month = {dec},
number = {S5},
title = {{Relating occlusion maps obtained through deep learning to functional impairment in dementia of Alzheimer's type}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/alz.043538},
volume = {16},
year = {2020}
}

@article{Santiago2020,
abstract = {The popularity of high and ultra-high definition displays has led to the need for methods to improve the quality of videos already obtained at much lower resolutions. A large amount of current CNN-based Video Super-Resolution methods are designed and trained to handle a specific degradation operator (e.g., bicubic downsampling) and are not robust to mismatch between training and testing degradation models. This causes their performance to deteriorate in real-life applications. Furthermore, many of them use the Mean-Squared-Error as the only loss during learning, causing the resulting images to be too smooth. In this work we propose a new Convolutional Neural Network for video super resolution which is robust to multiple degradation models. During training, which is performed on a large dataset of scenes with slow and fast motions, it uses the pseudo-inverse image formation model as part of the network architecture in conjunction with perceptual losses and a smoothness constraint that eliminates the artifacts originating from these perceptual losses. The experimental validation shows that our approach outperforms current state-of-the-art methods and is robust to multiple degradations.},
archivePrefix = {arXiv},
arxivId = {1907.01399},
author = {L{\'{o}}pez-Tapia, Santiago and Lucas, Alice and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.dsp.2020.102801},
eprint = {1907.01399},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Convolutional neuronal networks,Generative adversarial networks,Perceptual loss functions,Super-resolution,Video},
month = {sep},
pages = {102801},
title = {{A single video super-resolution GAN for multiple downsampling operators based on pseudo-inverse image formation models}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200420301469},
volume = {104},
year = {2020}
}

The popularity of high and ultra-high definition displays has led to the need for methods to improve the quality of videos already obtained at much lower resolutions. A large amount of current CNN-based Video Super-Resolution methods are designed and trained to handle a specific degradation operator (e.g., bicubic downsampling) and are not robust to mismatch between training and testing degradation models. This causes their performance to deteriorate in real-life applications. Furthermore, many of them use the Mean-Squared-Error as the only loss during learning, causing the resulting images to be too smooth. In this work we propose a new Convolutional Neural Network for video super resolution which is robust to multiple degradation models. During training, which is performed on a large dataset of scenes with slow and fast motions, it uses the pseudo-inverse image formation model as part of the network architecture in conjunction with perceptual losses and a smoothness constraint that eliminates the artifacts originating from these perceptual losses. The experimental validation shows that our approach outperforms current state-of-the-art methods and is robust to multiple degradations.

@article{Corey2020,
abstract = {Existing literature points to scaffolded training as an effective yet resource-intensive approach to help newcomers learn and stay motivated. Experts need to select relevant learning materials and continuously assess learners' progress. Peer production communities such as Wikipedia and Open Source Software Development projects face the additional problem of turning volunteers into productive participants as soon as possible. To address these challenges, we designed and tested a training regime combining scaffolded instruction and machine learning to select learning materials and gradually introduces new materials to individuals as their competences improve. We evaluated the training regime on 386 participants that contribute to Gravity Spy, an online citizen science project where people are asked to categorize glitches to assist scientists in the search for gravitational waves. Volunteers were assigned to one of two conditions; (1) a machine learning guided training (MLGT) system that continuously assesses volunteers skill level and adjusts the learning materials or (2) an unscaffolded training program where all learning materials were administered at once. Our analysis revealed that volunteers in the MLGT condition were more accurate on the categorization task (an average accuracy of 90% vs. 54%), executed more tasks (an average of 228 vs. 121 tasks), and were retained for a longer period (an average of 2.5 vs. 2 sessions) than volunteers in the unscaffolded training. The results suggest that MLGT is an effective pedagogical approach for training volunteers in categorization tasks and increases volunteers' motivation.},
author = {Jackson, Corey and {\O}sterlund, Carsten and Crowston, Kevin and Harandi, Mahboobeh and Allen, Sarah and Bahaadini, Sara and Coughlin, Scotty and Kalogera, Vicky and Katsaggelos, Aggelos and Larson, Shane and Rohani, Neda and Smith, Joshua and Trouille, Laura and Zevin, Michael},
doi = {10.1016/j.chb.2019.106198},
issn = {07475632},
journal = {Computers in Human Behavior},
keywords = {Citizen science,Experiment,Learning,Online communities,Scaffolding,Training,User studies,Zooniverse},
month = {apr},
pages = {106198},
title = {{Teaching citizen scientists to categorize glitches using machine learning guided training}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0747563219304182},
volume = {105},
year = {2020}
}

Existing literature points to scaffolded training as an effective yet resource-intensive approach to help newcomers learn and stay motivated. Experts need to select relevant learning materials and continuously assess learners' progress. Peer production communities such as Wikipedia and Open Source Software Development projects face the additional problem of turning volunteers into productive participants as soon as possible. To address these challenges, we designed and tested a training regime combining scaffolded instruction and machine learning to select learning materials and gradually introduces new materials to individuals as their competences improve. We evaluated the training regime on 386 participants that contribute to Gravity Spy, an online citizen science project where people are asked to categorize glitches to assist scientists in the search for gravitational waves. Volunteers were assigned to one of two conditions; (1) a machine learning guided training (MLGT) system that continuously assesses volunteers skill level and adjusts the learning materials or (2) an unscaffolded training program where all learning materials were administered at once. Our analysis revealed that volunteers in the MLGT condition were more accurate on the categorization task (an average accuracy of 90% vs. 54%), executed more tasks (an average of 228 vs. 121 tasks), and were retained for a longer period (an average of 2.5 vs. 2 sessions) than volunteers in the unscaffolded training. The results suggest that MLGT is an effective pedagogical approach for training volunteers in categorization tasks and increases volunteers' motivation.

@article{Qiqin2019,
abstract = {This paper presents an adaptive image sampling algorithm based on Deep Learning (DL). It consists of an adaptive sampling mask generation network which is jointly trained with an image inpainting network. The sampling rate is controlled by the mask generation network, and a binarization strategy is investigated to make the sampling mask binary. In addition to the image sampling and reconstruction process, we show how it can be extended and used to speed up raster scanning such as the X-Ray fluorescence (XRF) image scanning process. Recently XRF laboratory-based systems have evolved into lightweight and portable instruments thanks to technological advancements in both X-Ray generation and detection. However, the scanning time of an XRF image is usually long due to the long exposure requirements (e.g., 100\; $\mu$ s-1 ms per point). We propose an XRF image inpainting approach to address the long scanning times, thus speeding up the scanning process, while being able to reconstruct a high quality XRF image. The proposed adaptive image sampling algorithm is applied to the RGB image of the scanning target to generate the sampling mask. The XRF scanner is then driven according to the sampling mask to scan a subset of the total image pixels. Finally, we inpaint the scanned XRF image by fusing the RGB image to reconstruct the full scan XRF image. The experiments show that the proposed adaptive sampling algorithm is able to effectively sample the image and achieve a better reconstruction accuracy than that of existing methods.},
archivePrefix = {arXiv},
arxivId = {1812.10836},
author = {Dai, Qiqin and Chopp, Henry and Pouyet, Emeline and Cossairt, Oliver and Walton, Marc and Katsaggelos, Aggelos K.},
doi = {10.1109/TMM.2019.2958760},
eprint = {1812.10836},
issn = {1520-9210},
journal = {IEEE Transactions on Multimedia},
keywords = {Adaptive sampling,X-Ray fluorescence,convolutional neural network,inpainting},
month = {oct},
number = {10},
pages = {2564--2578},
title = {{Adaptive Image Sampling Using Deep Learning and Its Application on X-Ray Fluorescence Image Reconstruction}},
url = {https://ieeexplore.ieee.org/document/8930037/},
volume = {22},
year = {2020}
}

This paper presents an adaptive image sampling algorithm based on Deep Learning (DL). It consists of an adaptive sampling mask generation network which is jointly trained with an image inpainting network. The sampling rate is controlled by the mask generation network, and a binarization strategy is investigated to make the sampling mask binary. In addition to the image sampling and reconstruction process, we show how it can be extended and used to speed up raster scanning such as the X-Ray fluorescence (XRF) image scanning process. Recently XRF laboratory-based systems have evolved into lightweight and portable instruments thanks to technological advancements in both X-Ray generation and detection. However, the scanning time of an XRF image is usually long due to the long exposure requirements (e.g., 100  $μ$ s-1 ms per point). We propose an XRF image inpainting approach to address the long scanning times, thus speeding up the scanning process, while being able to reconstruct a high quality XRF image. The proposed adaptive image sampling algorithm is applied to the RGB image of the scanning target to generate the sampling mask. The XRF scanner is then driven according to the sampling mask to scan a subset of the total image pixels. Finally, we inpaint the scanned XRF image by fusing the RGB image to reconstruct the full scan XRF image. The experiments show that the proposed adaptive sampling algorithm is able to effectively sample the image and achieve a better reconstruction accuracy than that of existing methods.

@article{iliadis2020deepbinarymask,
abstract = {In this paper, we propose an encoder-decoder neural network model referred to as DeepBinaryMask for video compressive sensing. In video compressive sensing one frame is acquired using a set of coded masks (sensing matrix) from which a number of video frames, equal to the number of coded masks, is reconstructed. The proposed framework is an end-to-end model where the sensing matrix is trained along with the video reconstruction. The encoder maps a video block to compressive measurements by learning the binary elements of the sensing matrix. The decoder is trained to map the measurements from a video patch back to a video block via several hidden layers of a Multi-Layer Perceptron network. The predicted video blocks are stacked together to recover the unknown video sequence. The reconstruction performance is found to improve when using the trained sensing mask from the network as compared to other mask designs such as random, across a wide variety of compressive sensing reconstruction algorithms. Finally, our analysis and discussion offers insights into understanding the characteristics of the trained mask designs that lead to the improved reconstruction quality.},
archivePrefix = {arXiv},
arxivId = {1607.03343},
author = {Iliadis, Michael and Spinoulas, Leonidas and Katsaggelos, Aggelos K.},
doi = {10.1016/j.dsp.2019.102591},
eprint = {1607.03343},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Binary mask,Compressive sensing,Deep learning,Mask optimization,Video reconstruction},
month = {jan},
pages = {102591},
title = {{DeepBinaryMask: Learning a binary mask for video compressive sensing}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200419301459},
volume = {96},
year = {2020}
}

In this paper, we propose an encoder-decoder neural network model referred to as DeepBinaryMask for video compressive sensing. In video compressive sensing one frame is acquired using a set of coded masks (sensing matrix) from which a number of video frames, equal to the number of coded masks, is reconstructed. The proposed framework is an end-to-end model where the sensing matrix is trained along with the video reconstruction. The encoder maps a video block to compressive measurements by learning the binary elements of the sensing matrix. The decoder is trained to map the measurements from a video patch back to a video block via several hidden layers of a Multi-Layer Perceptron network. The predicted video blocks are stacked together to recover the unknown video sequence. The reconstruction performance is found to improve when using the trained sensing mask from the network as compared to other mask designs such as random, across a wide variety of compressive sensing reconstruction algorithms. Finally, our analysis and discussion offers insights into understanding the characteristics of the trained mask designs that lead to the improved reconstruction quality.

@article{Emily2020,
abstract = {Abstract Background Memory complaints are widespread among the elderly and aging is a major risk factor for Alzheimer's disease (AD), leading to the impression that gradual loss of memory ability is a nearly universal consequence of getting old. Our longitudinal studies of SuperAgers, 80+ year-olds with episodic memory performance that remains in the range that is at least normal for 50-60 year-olds suggests an alternative aging trajectory is possible. This session will highlight some of the emerging biologic features of the SuperAgers. Method Participants include SuperAgers and cognitively average 80+ year-old cognitively average normal controls. Data from detailed neuropsychological assessments, quantitative neuroimaging measurements (MR and amyloid PET), genetic features and neuropathologic findings will be reported. Result Initial evidence suggest SuperAgers tend to show mismatch between chronologic and biologic age, including maintenance of cortical but not necessarily hippocampal volume, a tendency to resist significant amyloid PET retention, an abundance of anterior cingulate Von Economo neurons, and some with resistance to cortical Alzheimer's pathology. Conclusion These studies contribute to our understanding mechanisms of resilience and resistance in cognitive aging and may help isolate factors that are potentially important for promoting successful cognitive aging and avoiding age-related brain diseases such as AD.},
author = {Rogalski, Emily J and Sridhar, Jaiashre and Martersteck, Adam and Makowski‐Woidan, Beth and Engelmeyer, Janessa and Parrish, Todd and Besson, Pierre and Cobia, Derin and Paxton, Holly and Weintraub, Sandra and Katsaggelos, Aggelos and Bandt, Katie and Bigio, Eileen H and Geula, Changiz and Mesulam, Marsel},
doi = {10.1002/alz.037932},
issn = {1552-5260},
journal = {Alzheimer's & Dementia},
month = {dec},
number = {S10},
pages = {e037932},
title = {{SuperAging: A model for studying mechanisms of resilience and resistance}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/alz.037932},
volume = {16},
year = {2020}
}

Abstract Background Memory complaints are widespread among the elderly and aging is a major risk factor for Alzheimer's disease (AD), leading to the impression that gradual loss of memory ability is a nearly universal consequence of getting old. Our longitudinal studies of SuperAgers, 80+ year-olds with episodic memory performance that remains in the range that is at least normal for 50-60 year-olds suggests an alternative aging trajectory is possible. This session will highlight some of the emerging biologic features of the SuperAgers. Method Participants include SuperAgers and cognitively average 80+ year-old cognitively average normal controls. Data from detailed neuropsychological assessments, quantitative neuroimaging measurements (MR and amyloid PET), genetic features and neuropathologic findings will be reported. Result Initial evidence suggest SuperAgers tend to show mismatch between chronologic and biologic age, including maintenance of cortical but not necessarily hippocampal volume, a tendency to resist significant amyloid PET retention, an abundance of anterior cingulate Von Economo neurons, and some with resistance to cortical Alzheimer's pathology. Conclusion These studies contribute to our understanding mechanisms of resilience and resistance in cognitive aging and may help isolate factors that are potentially important for promoting successful cognitive aging and avoiding age-related brain diseases such as AD.

@article{Haoyu2020,
abstract = {Computed tomography is widely used to examine internal structures in a non-destructive manner. To obtain high-quality reconstructions, one typically has to acquire a densely sampled trajectory to avoid angular undersampling. However, many scenarios require a sparse-view measurement leading to streak-artifacts if unaccounted for. Current methods do not make full use of the domain-specific information, and hence fail to provide reliable reconstructions for highly undersampled data. We present a novel framework for sparse-view tomography by decoupling the reconstruction into two steps: First, we overcome its ill-posedness using a super-resolution network, SIN, trained on the sparse projections. The intermediate result allows for a closed-form tomographic reconstruction with preserved details and highly reduced streak-artifacts. Second, a refinement network, PRN, trained on the reconstructions reduces any remaining artifacts. We further propose a light-weight variant of the perceptual-loss that enhances domain-specific information, boosting restoration accuracy. Our experiments demonstrate an improvement over current solutions by 4 dB.},
archivePrefix = {arXiv},
arxivId = {2012.04743},
author = {Wei, Haoyu and Schiffers, Florian and W{\"{u}}rfl, Tobias and Shen, Daming and Kim, Daniel and Katsaggelos, Aggelos K. and Cossairt, Oliver},
eprint = {2012.04743},
journal = {arXiv preprint arXiv:2012.04743},
month = {dec},
title = {{2-Step Sparse-View CT Reconstruction with a Domain-Specific Perceptual Network}},
url = {http://arxiv.org/abs/2012.04743},
year = {2020}
}

Computed tomography is widely used to examine internal structures in a non-destructive manner. To obtain high-quality reconstructions, one typically has to acquire a densely sampled trajectory to avoid angular undersampling. However, many scenarios require a sparse-view measurement leading to streak-artifacts if unaccounted for. Current methods do not make full use of the domain-specific information, and hence fail to provide reliable reconstructions for highly undersampled data. We present a novel framework for sparse-view tomography by decoupling the reconstruction into two steps: First, we overcome its ill-posedness using a super-resolution network, SIN, trained on the sparse projections. The intermediate result allows for a closed-form tomographic reconstruction with preserved details and highly reduced streak-artifacts. Second, a refinement network, PRN, trained on the reconstructions reduces any remaining artifacts. We further propose a light-weight variant of the perceptual-loss that enhances domain-specific information, boosting restoration accuracy. Our experiments demonstrate an improvement over current solutions by 4 dB.

@article{Arjun2019,
abstract = {Automated methods for Alzheimer's disease (AD) classification have the potential for great clinical benefits and may provide insight for combating the disease. Machine learning, and more specifically deep neural networks, have been shown to have great efficacy in this domain. These algorithms often use neurological imaging data such as MRI and FDG PET, but a comprehensive and balanced comparison of the MRI and amyloid PET modalities has not been performed. In order to accurately determine the relative strength of each imaging variant, this work performs a comparison study in the context of Alzheimer's dementia classification using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset with identical neural network architectures. Furthermore, this work analyzes the benefits of using both modalities in a fusion setting and discusses how these data types may be leveraged in future AD studies using deep learning.},
archivePrefix = {arXiv},
arxivId = {1811.05105},
author = {Punjabi, Arjun and Martersteck, Adam and Wang, Yanran and Parrish, Todd B. and Katsaggelos, Aggelos K.},
doi = {10.1371/journal.pone.0225759},
editor = {Ginsberg, Stephen D},
eprint = {1811.05105},
issn = {1932-6203},
journal = {PLOS ONE},
month = {dec},
number = {12},
pages = {e0225759},
pmid = {31805160},
title = {{Neuroimaging modality fusion in Alzheimer's classification using convolutional neural networks}},
url = {https://dx.plos.org/10.1371/journal.pone.0225759},
volume = {14},
year = {2019}
}

Automated methods for Alzheimer's disease (AD) classification have the potential for great clinical benefits and may provide insight for combating the disease. Machine learning, and more specifically deep neural networks, have been shown to have great efficacy in this domain. These algorithms often use neurological imaging data such as MRI and FDG PET, but a comprehensive and balanced comparison of the MRI and amyloid PET modalities has not been performed. In order to accurately determine the relative strength of each imaging variant, this work performs a comparison study in the context of Alzheimer's dementia classification using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset with identical neural network architectures. Furthermore, this work analyzes the benefits of using both modalities in a fusion setting and discusses how these data types may be leveraged in future AD studies using deep learning.

@article{AliBahrami2019,
abstract = {Objective: The aim of this study is to develop an automated classification algorithm for polysomnography (PSG) recordings to detect non-apneic and non-hypopneic arousals. Our particular focus is on detecting the respiratory effort-related arousals (RERAs) which are very subtle respiratory events that do not meet the criteria for apnea or hypopnea, and are more challenging to detect. Methods: The proposed algorithm is based on a bidirectional long short-term memory (BiLSTM) classifier and 465 multi-domain features, extracted from multimodal clinical time series. The features consist of a set of physiology-inspired features (n = 75), obtained by multiple steps of feature selection and expert analysis, and a set of physiology-agnostic features (n = 390), derived from scattering transform. Results: The proposed algorithm is validated on the 2018 PhysioNet challenge dataset. The overall performance in terms of the area under the precision-recall curve (AUPRC) is 0.50 on the hidden test dataset. This result is tied for the second-best score during the follow-up and official phases of the 2018 PhysioNet challenge. Conclusions: The results demonstrate that it is possible to automatically detect subtle non-apneic/non-hypopneic arousal events from PSG recordings. Significance: Automatic detection of subtle respiratory events such as RERAs together with other non-apneic/non-hypopneic arousals will allow detailed annotations of large PSG databases. This contributes to a better retrospective analysis of sleep data, which may also improve the quality of treatment.},
archivePrefix = {arXiv},
arxivId = {1909.02971},
author = {Rad, Ali Bahrami and Zabihi, Morteza and Zhao, Zheng and Gabbouj, Moncef and Katsaggelos, Aggelos K. and S{\"{a}}rkk{\"{a}}, Simo},
eprint = {1909.02971},
journal = {arXiv e-prints},
month = {sep},
pages = {arXiv----1909},
title = {{Automated Polysomnography Analysis for Detection of Non-Apneic and Non-Hypopneic Arousals using Feature Engineering and a Bidirectional LSTM Network}},
url = {http://arxiv.org/abs/1909.02971},
year = {2019}
}

Objective: The aim of this study is to develop an automated classification algorithm for polysomnography (PSG) recordings to detect non-apneic and non-hypopneic arousals. Our particular focus is on detecting the respiratory effort-related arousals (RERAs) which are very subtle respiratory events that do not meet the criteria for apnea or hypopnea, and are more challenging to detect. Methods: The proposed algorithm is based on a bidirectional long short-term memory (BiLSTM) classifier and 465 multi-domain features, extracted from multimodal clinical time series. The features consist of a set of physiology-inspired features (n = 75), obtained by multiple steps of feature selection and expert analysis, and a set of physiology-agnostic features (n = 390), derived from scattering transform. Results: The proposed algorithm is validated on the 2018 PhysioNet challenge dataset. The overall performance in terms of the area under the precision-recall curve (AUPRC) is 0.50 on the hidden test dataset. This result is tied for the second-best score during the follow-up and official phases of the 2018 PhysioNet challenge. Conclusions: The results demonstrate that it is possible to automatically detect subtle non-apneic/non-hypopneic arousal events from PSG recordings. Significance: Automatic detection of subtle respiratory events such as RERAs together with other non-apneic/non-hypopneic arousals will allow detailed annotations of large PSG databases. This contributes to a better retrospective analysis of sleep data, which may also improve the quality of treatment.

@article{coughlin2019classifying,
abstract = {The observation of gravitational waves from compact binary coalescences by LIGO and Virgo has begun a new era in astronomy. A critical challenge in making detections is determining whether loud transient features in the data are caused by gravitational waves or by instrumental or environmental sources. The citizen-science project Gravity Spy has been demonstrated as an efficient infrastructure for classifying known types of noise transients (glitches) through a combination of data analysis performed by both citizen volunteers and machine learning. We present the next iteration of this project, using similarity indices to empower citizen scientists to create large data sets of unknown transients, which can then be used to facilitate supervised machine-learning characterization. This new evolution aims to alleviate a persistent challenge that plagues both citizen-science and instrumental detector work: the ability to build large samples of relatively rare events. Using two families of transient noise that appeared unexpectedly during LIGO's second observing run, we demonstrate the impact that the similarity indices could have had on finding these new glitch types in the Gravity Spy program.},
archivePrefix = {arXiv},
arxivId = {1903.04058},
author = {Coughlin, S. and Bahaadini, S. and Rohani, N. and Zevin, M. and Patane, O. and Harandi, M. and Jackson, C. and Noroozi, V. and Allen, S. and Areeda, J. and Coughlin, M. and Ruiz, P. and Berry, C. P. L. and Crowston, K. and Katsaggelos, A. K. and Lundgren, A. and {\O}sterlund, C. and Smith, J. R. and Trouille, L. and Kalogera, V.},
doi = {10.1103/PhysRevD.99.082002},
eprint = {1903.04058},
issn = {2470-0010},
journal = {Physical Review D},
month = {apr},
number = {8},
pages = {082002},
publisher = {American Physical Society},
title = {{Classifying the unknown: Discovering novel gravitational-wave detector glitches using similarity learning}},
url = {https://link.aps.org/doi/10.1103/PhysRevD.99.082002},
volume = {99},
year = {2019}
}

The observation of gravitational waves from compact binary coalescences by LIGO and Virgo has begun a new era in astronomy. A critical challenge in making detections is determining whether loud transient features in the data are caused by gravitational waves or by instrumental or environmental sources. The citizen-science project Gravity Spy has been demonstrated as an efficient infrastructure for classifying known types of noise transients (glitches) through a combination of data analysis performed by both citizen volunteers and machine learning. We present the next iteration of this project, using similarity indices to empower citizen scientists to create large data sets of unknown transients, which can then be used to facilitate supervised machine-learning characterization. This new evolution aims to alleviate a persistent challenge that plagues both citizen-science and instrumental detector work: the ability to build large samples of relatively rare events. Using two families of transient noise that appeared unexpectedly during LIGO's second observing run, we demonstrate the impact that the similarity indices could have had on finding these new glitch types in the Gravity Spy program.

@article{Alice2019a,
abstract = {While Convolutional Neural Networks (CNNs) trained for image and video super-resolution (SR) regularly achieve new state-of-the-art performance, they also suffer from significant drawbacks. One of their limitations is their lack of robustness to unseen image formation models during training. Other limitations include the generation of artifacts and hallucinated content when training Generative Adversarial Networks (GANs) for SR. While the Deep Learning literature focuses on presenting new training schemes and settings to resolve these various issues, we show that one can avoid training and correct for SR results with a fully self-supervised fine-tuning approach. More specifically, at test time, given an image and its known image formation model, we fine-tune the parameters of the trained network and iteratively update them using a data fidelity loss. We apply our fine-tuning algorithm on multiple image and video SR CNNs and show that it can successfully correct for a sub-optimal SR solution by entirely relying on internal learning at test time. We apply our method on the problem of fine-tuning for unseen image formation models and on removal of artifacts introduced by GANs.},
archivePrefix = {arXiv},
arxivId = {https://arxiv.org/abs/1912.12879v3},
author = {Lucas, Alice and Lopez-Tapia, Santiago and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {https://doi.org/10.48550/arXiv.1912.12879},
eprint = {/arxiv.org/abs/1912.12879v3},
journal = {arXiv preprint arXiv:1912.12879},
keywords = {Computer Science - Machine Learning,Electrical Engineering and Systems Science - Image,Statistics - Machine Learning},
month = {dec},
pages = {1--15},
primaryClass = {https:},
title = {{Self-supervised Fine-tuning for Correcting Super-Resolution Convolutional Neural Networks}},
url = {http://arxiv.org/abs/1912.12879},
year = {2019}
}

While Convolutional Neural Networks (CNNs) trained for image and video super-resolution (SR) regularly achieve new state-of-the-art performance, they also suffer from significant drawbacks. One of their limitations is their lack of robustness to unseen image formation models during training. Other limitations include the generation of artifacts and hallucinated content when training Generative Adversarial Networks (GANs) for SR. While the Deep Learning literature focuses on presenting new training schemes and settings to resolve these various issues, we show that one can avoid training and correct for SR results with a fully self-supervised fine-tuning approach. More specifically, at test time, given an image and its known image formation model, we fine-tune the parameters of the trained network and iteratively update them using a data fidelity loss. We apply our fine-tuning algorithm on multiple image and video SR CNNs and show that it can successfully correct for a sub-optimal SR solution by entirely relying on internal learning at test time. We apply our method on the problem of fine-tuning for unseen image formation models and on removal of artifacts introduced by GANs.

@article{Alice2019c,
abstract = {Video super-resolution (VSR) has become one of the most critical problems in video processing. In the deep learning literature, recent works have shown the benefits of using adversarial-based and perceptual losses to improve the performance on various image restoration tasks; however, these have yet to be applied for video super-resolution. In this paper, we propose a generative adversarial network (GAN)-based formulation for VSR. We introduce a new generator network optimized for the VSR problem, named VSRResNet, along with new discriminator architecture to properly guide VSRResNet during the GAN training. We further enhance our VSR GAN formulation with two regularizers, a distance loss in feature-space and pixel-space, to obtain our final VSRResFeatGAN model. We show that pre-training our generator with the mean-squared-error loss only quantitatively surpasses the current state-of-the-art VSR models. Finally, we employ the PercepDist metric to compare the state-of-the-art VSR models. We show that this metric more accurately evaluates the perceptual quality of SR solutions obtained from neural networks, compared with the commonly used PSNR/SSIM metrics. Finally, we show that our proposed model, the VSRResFeatGAN model, outperforms the current state-of-the-art SR models, both quantitatively and qualitatively.},
archivePrefix = {arXiv},
arxivId = {1806.05764},
author = {Lucas, Alice and Lopez-Tapia, Santiago and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2019.2895768},
eprint = {1806.05764},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Artificial neural networks,image generation,image resolution,video signal processing},
month = {jul},
number = {7},
pages = {3312--3327},
pmid = {30714918},
title = {{Generative Adversarial Networks and Perceptual Losses for Video Super-Resolution}},
url = {https://ieeexplore.ieee.org/document/8629024/},
volume = {28},
year = {2019}
}

Video super-resolution (VSR) has become one of the most critical problems in video processing. In the deep learning literature, recent works have shown the benefits of using adversarial-based and perceptual losses to improve the performance on various image restoration tasks; however, these have yet to be applied for video super-resolution. In this paper, we propose a generative adversarial network (GAN)-based formulation for VSR. We introduce a new generator network optimized for the VSR problem, named VSRResNet, along with new discriminator architecture to properly guide VSRResNet during the GAN training. We further enhance our VSR GAN formulation with two regularizers, a distance loss in feature-space and pixel-space, to obtain our final VSRResFeatGAN model. We show that pre-training our generator with the mean-squared-error loss only quantitatively surpasses the current state-of-the-art VSR models. Finally, we employ the PercepDist metric to compare the state-of-the-art VSR models. We show that this metric more accurately evaluates the perceptual quality of SR solutions obtained from neural networks, compared with the commonly used PSNR/SSIM metrics. Finally, we show that our proposed model, the VSRResFeatGAN model, outperforms the current state-of-the-art SR models, both quantitatively and qualitatively.

@article{Arun2019,
abstract = {A novel optical flow prediction model using an adaptable deep neural network architecture for blind and non-blind error concealment of videos degraded by transmission loss is presented. The two-stream network model is trained by separating the horizontal and vertical motion fields which are passed through two similar parallel pipelines that include traditional convolutional (Conv) and convolutional long short-term memory (ConvLSTM) layers. The ConvLSTM layers extract temporally correlated motion information while the Conv layers correlate motion spatially. The optical flows used as input to the two-pipeline prediction network are obtained through a flow generation network that can be easily interchanged, increasing the adaptability of the overall end-to-end architecture. The performance of the proposed model is evaluated using real-world packet loss scenarios. Standard video quality metrics are used to compare frames reconstructed using predicted optical flows with those reconstructed using “ground-truth” flows obtained directly from the generator.},
author = {Sankisa, Arun and Punjabi, Arjun and Katsaggelos, Aggelos K.},
doi = {10.4018/IJMDEM.2019070102},
isbn = {978-1-4799-7061-2},
issn = {1947-8534},
journal = {International Journal of Multimedia Data Engineering and Management},
keywords = {CNN,ConvLSTM,Deep Neural Networks,Optical flow,Video Error Concealment},
month = {jul},
number = {3},
pages = {27--46},
publisher = {IEEE},
title = {{Optical Flow Prediction for Blind and Non-Blind Video Error Concealment Using Deep Neural Networks}},
url = {https://ieeexplore.ieee.org/document/8451090/ http://services.igi-global.com/resolvedoi/resolve.aspx?doi=10.4018/IJMDEM.2019070102},
volume = {10},
year = {2019}
}

A novel optical flow prediction model using an adaptable deep neural network architecture for blind and non-blind error concealment of videos degraded by transmission loss is presented. The two-stream network model is trained by separating the horizontal and vertical motion fields which are passed through two similar parallel pipelines that include traditional convolutional (Conv) and convolutional long short-term memory (ConvLSTM) layers. The ConvLSTM layers extract temporally correlated motion information while the Conv layers correlate motion spatially. The optical flows used as input to the two-pipeline prediction network are obtained through a flow generation network that can be easily interchanged, increasing the adaptability of the overall end-to-end architecture. The performance of the proposed model is evaluated using real-world packet loss scenarios. Standard video quality metrics are used to compare frames reconstructed using predicted optical flows with those reconstructed using “ground-truth” flows obtained directly from the generator.

@article{Pablo2019,
abstract = {Over the last few years, multiply-annotated data has become a very popular source of information. Online platforms such as Amazon Mechanical Turk have revolutionized the labelling process needed for any classification task, sharing the effort between a number of annotators (instead of the classical single expert). This crowdsourcing approach has introduced new challenging problems, such as handling disagreements on the annotated samples or combining the unknown expertise of the annotators. Probabilistic methods, such as Gaussian Processes (GP), have proven successful to model this new crowdsourcing scenario. However, GPs do not scale up well with the training set size, which makes them prohibitive for medium-to-large datasets (beyond 10K training instances). This constitutes a serious limitation for current real-world applications. In this work, we introduce two scalable and efficient GP-based crowdsourcing methods that allow for processing previously-prohibitive datasets. The first one is an efficient and fast approximation to GP with squared exponential (SE) kernel. The second allows for learning a more flexible kernel at the expense of a heavier training (but still scalable to large datasets). Since the latter is not a GP-SE approximation, it can be also considered as a whole new scalable and efficient crowdsourcing method, useful for any dataset size. Both methods use Fourier features and variational inference, can predict the class of new samples, and estimate the expertise of the involved annotators. A complete experimentation compares them with state-of-the-art probabilistic approaches in synthetic and real crowdsourcing datasets of different sizes. They stand out as the best performing approach for large scale problems. Moreover, the second method is competitive with the current state-of-the-art for small datasets.},
author = {Morales-{\'{A}}lvarez, Pablo and Ruiz, Pablo and Santos-Rodr{\'{i}}guez, Ra{\'{u}}l and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.inffus.2018.12.008},
issn = {15662535},
journal = {Information Fusion},
keywords = {Bayesian modelling,Classification,Fourier features,Gaussian processes,Scalable crowdsourcing,Variational inference},
month = {dec},
pages = {110--127},
title = {{Scalable and efficient learning from crowds with Gaussian processes}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1566253518304664},
volume = {52},
year = {2019}
}

Over the last few years, multiply-annotated data has become a very popular source of information. Online platforms such as Amazon Mechanical Turk have revolutionized the labelling process needed for any classification task, sharing the effort between a number of annotators (instead of the classical single expert). This crowdsourcing approach has introduced new challenging problems, such as handling disagreements on the annotated samples or combining the unknown expertise of the annotators. Probabilistic methods, such as Gaussian Processes (GP), have proven successful to model this new crowdsourcing scenario. However, GPs do not scale up well with the training set size, which makes them prohibitive for medium-to-large datasets (beyond 10K training instances). This constitutes a serious limitation for current real-world applications. In this work, we introduce two scalable and efficient GP-based crowdsourcing methods that allow for processing previously-prohibitive datasets. The first one is an efficient and fast approximation to GP with squared exponential (SE) kernel. The second allows for learning a more flexible kernel at the expense of a heavier training (but still scalable to large datasets). Since the latter is not a GP-SE approximation, it can be also considered as a whole new scalable and efficient crowdsourcing method, useful for any dataset size. Both methods use Fourier features and variational inference, can predict the class of new samples, and estimate the expertise of the involved annotators. A complete experimentation compares them with state-of-the-art probabilistic approaches in synthetic and real crowdsourcing datasets of different sizes. They stand out as the best performing approach for large scale problems. Moreover, the second method is competitive with the current state-of-the-art for small datasets.

@article{Alice2019,
abstract = {Somatosensation is composed of two distinct modalities: touch, arising from sensors in the skin, and proprioception, resulting primarily from sensors in the muscles, combined with these same cutaneous sensors. In contrast to the wealth of information about touch, we know quite less about the nature of the signals giving rise to proprioception at the cortical level. Likewise, while there is considerable interest in developing encoding models of touch-related neurons for application to brain machine interfaces, much less emphasis has been placed on an analogous proprioceptive interface. Here we investigate the use of Artificial Neural Networks (ANNs) to model the relationship between the firing rates of single neurons in area 2, a largely proprioceptive region of somatosensory cortex (S1) and several types of kinematic variables related to arm movement. To gain a better understanding of how these kinematic variables interact to create the proprioceptive responses recorded in our datasets, we train ANNs under different conditions, each involving a different set of input and output variables. We explore the kinematic variables that provide the best network performance, and find that the addition of information about joint angles and/or muscle lengths significantly improves the prediction of neural firing rates. Our results thus provide new insight regarding the complex representations of the limb motion in S1: that the firing rates of neurons in area 2 may be more closely related to the activity of peripheral sensors than it is to extrinsic hand position. In addition, we conduct numerical experiments to determine the sensitivity of ANN models to various choices of training design and hyper-parameters. Our results provide a baseline and new tools for future research that utilizes machine learning to better describe and understand the activity of neurons in S1.},
author = {Lucas, Alice and Tomlinson, Tucker and Rohani, Neda and Chowdhury, Raeed and Solla, Sara A. and Katsaggelos, Aggelos K. and Miller, Lee E.},
doi = {10.3389/fnsys.2019.00013},
issn = {1662-5137},
journal = {Frontiers in Systems Neuroscience},
keywords = {Artificial neural networks,Limb-state encoding,Monkey,Reaching,Single neurons,Somatosensory cortex},
month = {mar},
pages = {13 , publisher = Frontiers Media SA},
title = {{Neural Networks for Modeling Neural Spiking in S1 Cortex}},
url = {https://www.frontiersin.org/article/10.3389/fnsys.2019.00013/full},
volume = {13},
year = {2019}
}

Somatosensation is composed of two distinct modalities: touch, arising from sensors in the skin, and proprioception, resulting primarily from sensors in the muscles, combined with these same cutaneous sensors. In contrast to the wealth of information about touch, we know quite less about the nature of the signals giving rise to proprioception at the cortical level. Likewise, while there is considerable interest in developing encoding models of touch-related neurons for application to brain machine interfaces, much less emphasis has been placed on an analogous proprioceptive interface. Here we investigate the use of Artificial Neural Networks (ANNs) to model the relationship between the firing rates of single neurons in area 2, a largely proprioceptive region of somatosensory cortex (S1) and several types of kinematic variables related to arm movement. To gain a better understanding of how these kinematic variables interact to create the proprioceptive responses recorded in our datasets, we train ANNs under different conditions, each involving a different set of input and output variables. We explore the kinematic variables that provide the best network performance, and find that the addition of information about joint angles and/or muscle lengths significantly improves the prediction of neural firing rates. Our results thus provide new insight regarding the complex representations of the limb motion in S1: that the firing rates of neurons in area 2 may be more closely related to the activity of peripheral sensors than it is to extrinsic hand position. In addition, we conduct numerical experiments to determine the sensitivity of ANN models to various choices of training design and hyper-parameters. Our results provide a baseline and new tools for future research that utilizes machine learning to better describe and understand the activity of neurons in S1.

@article{Pablo2019a,
abstract = {Historical paper often contains features embedded in its structure that are invisible under standard viewing conditions. These features (watermarks, laid lines, and chain lines) can provide valuable information about a sheet's provenance. Standard methods of reproducing watermarks, such as beta-radiography and low-voltage x-rays, are costly and time intensive, and therefore inaccessible to many institutions or individuals. In this work we introduce an inexpensive prototype whose elements are a light table and a consumer-grade photographic camera. For a given document we acquire an image with light emitted by a light table passing through the document and two images of the front and the back side with ambient light. The images are then processed to suppress the printed elements and isolate the watermark. The proposed method is capable of recovering images of watermarks similar to the ones obtained with standard methods while being a non-destructive, rapid, easy to operate, and inexpensive method.},
author = {Ruiz, Pablo and Dill, Olivia and Raju, Goutam and Cossairt, Oliver and Walton, Marc and Katsaggelos, Aggelos K.},
doi = {10.1016/j.daach.2019.e00121},
issn = {22120548},
journal = {Digital Applications in Archaeology and Cultural Heritage},
keywords = {Image processing,Registration,Scattering,Visible light,Watermark extraction},
month = {dec},
pages = {e00121},
title = {{Visible transmission imaging of watermarks by suppression of occluding text or drawings}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S221205481830047X},
volume = {15},
year = {2019}
}

Historical paper often contains features embedded in its structure that are invisible under standard viewing conditions. These features (watermarks, laid lines, and chain lines) can provide valuable information about a sheet's provenance. Standard methods of reproducing watermarks, such as beta-radiography and low-voltage x-rays, are costly and time intensive, and therefore inaccessible to many institutions or individuals. In this work we introduce an inexpensive prototype whose elements are a light table and a consumer-grade photographic camera. For a given document we acquire an image with light emitted by a light table passing through the document and two images of the front and the back side with ambient light. The images are then processed to suppress the printed elements and isolate the watermark. The proposed method is capable of recovering images of watermarks similar to the ones obtained with standard methods while being a non-destructive, rapid, easy to operate, and inexpensive method.

@article{Juan2019,
abstract = {Most state-of-the-art blind image deconvolution methods rely on the Bayesian paradigm to model the deblurring problem and estimate both the blur kernel and latent image. It is customary to model the image in the filter space, where it is supposed to be sparse, and utilize convenient priors to account for this sparsity. In this paper, we propose the use of the spike-and-slab prior together with an efficient variational Expectation Maximization (EM) inference scheme to estimate the blur in the image. The spike-and-slab prior, which constitutes the gold standard in sparse machine learning, selectively shrinks irrelevant variables while mildly regularizing the relevant ones. The proposed variational Expectation Maximization algorithm is more efficient than usual Markov Chain Monte Carlo (MCMC) inference and, also, proves to be more accurate than the standard mean-field variational approximation. Additionally, all the prior model parameters are estimated by the proposed scheme. After blur estimation, a non-blind restoration method is used to obtain the actual estimation of the sharp image. We investigate the behavior of the prior in the experimental section together with a series of experiments with synthetically generated and real blurred images that validate the method's performance in comparison with state-of-the-art blind deconvolution techniques.},
author = {Serra, Juan G. and Mateos, Javier and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.dsp.2019.01.004},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Blind image deconvolution,Blur estimation,Spike and slab,Variational inference},
month = {may},
pages = {116--129},
title = {{Variational EM method for blur estimation using the spike-and-slab image prior}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200418306109},
volume = {88},
year = {2019}
}

Most state-of-the-art blind image deconvolution methods rely on the Bayesian paradigm to model the deblurring problem and estimate both the blur kernel and latent image. It is customary to model the image in the filter space, where it is supposed to be sparse, and utilize convenient priors to account for this sparsity. In this paper, we propose the use of the spike-and-slab prior together with an efficient variational Expectation Maximization (EM) inference scheme to estimate the blur in the image. The spike-and-slab prior, which constitutes the gold standard in sparse machine learning, selectively shrinks irrelevant variables while mildly regularizing the relevant ones. The proposed variational Expectation Maximization algorithm is more efficient than usual Markov Chain Monte Carlo (MCMC) inference and, also, proves to be more accurate than the standard mean-field variational approximation. Additionally, all the prior model parameters are estimated by the proposed scheme. After blur estimation, a non-blind restoration method is used to obtain the actual estimation of the sharp image. We investigate the behavior of the prior in the experimental section together with a series of experiments with synthetically generated and real blurred images that validate the method's performance in comparison with state-of-the-art blind deconvolution techniques.

@article{Neda2015,
abstract = {This paper proposes a new model for multi-sensory data classification. To tackle this problem, probabilistic modeling and variational Bayesian inference are used. A Gaussian Process (GP) classifier is built upon the introduced modeling. Its posterior distribution is approximated using variational Bayesian inference. Finally, labels of test samples are predicted employing this classifier. Very importantly, and in contrast to alternative approaches, the proposed method does not discard samples with missing features and utilizes all available information for training. Furthermore, to take into account that the quality of the information provided by each sensor may differ (some modalities/sensors may provide more reliable/distinctive information than others), we introduce two versions of the algorithm. In the first one, the parameters modeling each sensor performance are shared while in the second one, each sensor parameters are estimated independently. Synthetic and real datasets are utilized to examine the validity of the proposed models. The results obtained for binary classification problems justify their use and confirm their superiority over existing fusion architectures.},
author = {Rohani, Neda and Ruiz, Pablo and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.patrec.2018.08.035},
issn = {01678655},
journal = {Pattern Recognition Letters},
keywords = {Fusion,Gaussian process,Kernel,Posterior probability,Variational inference},
month = {dec},
pages = {80--87},
title = {{Variational Gaussian process for multisensor classification problems}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0167865518305427},
volume = {116},
year = {2018}
}

This paper proposes a new model for multi-sensory data classification. To tackle this problem, probabilistic modeling and variational Bayesian inference are used. A Gaussian Process (GP) classifier is built upon the introduced modeling. Its posterior distribution is approximated using variational Bayesian inference. Finally, labels of test samples are predicted employing this classifier. Very importantly, and in contrast to alternative approaches, the proposed method does not discard samples with missing features and utilizes all available information for training. Furthermore, to take into account that the quality of the information provided by each sensor may differ (some modalities/sensors may provide more reliable/distinctive information than others), we introduce two versions of the algorithm. In the first one, the parameters modeling each sensor performance are shared while in the second one, each sensor parameters are estimated independently. Synthetic and real datasets are utilized to examine the validity of the proposed models. The results obtained for binary classification problems justify their use and confirm their superiority over existing fusion architectures.

@article{Salvador2018,
abstract = {Digital Forensics encompasses the recovery and investigation of data, images, and recordings found in digital devices in order to provide evidence in the court of law. This paper is devoted to the assessment of digital evidence which requires not only an understanding of the scientific technique that leads to improved quality of surveillance video recordings, but also of the legal principles behind it. Emphasis is given on the special treatment of image processing in terms of its handling and explanation that would be acceptable in a court of law. In this context, we propose a variational Bayesian approach to multiple-image super-resolution based on Super-Gaussian prior models that automatically enhances the quality of outdoor video recordings and estimates all the model parameters while preserving the authenticity, credibility and reliability of video data as digital evidence. The proposed methodology is validated both quantitatively and visually on synthetic videos generated from single images and real-life videos and applied to a real-life case of damages and stealing in a private property.},
author = {Villena, Salvador and Vega, Miguel and Mateos, Javier and Rosenberg, Duska and Murtagh, Fionn and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.compind.2018.02.004},
issn = {01663615},
journal = {Computers in Industry},
keywords = {Legal aspects,Outdoor surveillance,Super-resolution,Usability},
month = {jun},
pages = {34--47},
title = {{Image super-resolution for outdoor digital forensics. Usability and legal aspects}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0166361517306358},
volume = {98},
year = {2018}
}

Digital Forensics encompasses the recovery and investigation of data, images, and recordings found in digital devices in order to provide evidence in the court of law. This paper is devoted to the assessment of digital evidence which requires not only an understanding of the scientific technique that leads to improved quality of surveillance video recordings, but also of the legal principles behind it. Emphasis is given on the special treatment of image processing in terms of its handling and explanation that would be acceptable in a court of law. In this context, we propose a variational Bayesian approach to multiple-image super-resolution based on Super-Gaussian prior models that automatically enhances the quality of outdoor video recordings and estimates all the model parameters while preserving the authenticity, credibility and reliability of video data as digital evidence. The proposed methodology is validated both quantitatively and visually on synthetic videos generated from single images and real-life videos and applied to a real-life case of damages and stealing in a private property.

@article{Xiang2018a,
abstract = {Volumetric biological imaging often involves compromising high temporal resolution at the expense of high spatial resolution when popular scanning methods are used to capture 3D information. We introduce an integrated experimental and image reconstruction method for capturing dynamic 3D fluorescent extended objects as a series of synchronously measured 3D snapshots taken at the frame rate of the imaging camera. We employ multifocal microscopy (MFM) to simultaneously image at 25 focal planes and process this depth-encoded image to recover the 3D structure of extended objects, such as bacteria, using a sparsity-based reconstruction approach. The combined experimental and computational method produces image quality similar to confocal microscopy in a fraction of the acquisition time. In addition, our computational image reconstruction approach allows a simplified MFM optical design by correcting aberrations using the measured response to point sources. This "compressive" MFM acquisition and reconstruction method, where an image volume with roughly 8 million voxels is recovered from a single 1-megapixel captured image, enables straightforward study of dynamic processes in 3D, and as a simultaneous snapshot advances the state of the art in dynamic 3D microscopy.},
archivePrefix = {arXiv},
arxivId = {1802.01565},
author = {Huang, Xiang and Selewa, Alan and Wang, Xiaolei and Daddysman, Matthew K. and Gdor, Itay and Wilton, Rosemarie and Kemner, Kenneth M. and Yoo, Seunghwan and Katsaggelos, Aggelos K. and He, Kuan and Cossairt, Oliver and Ferrier, Nicola J. and Hereld, Mark and Scherer, Norbert F.},
eprint = {1802.01565},
journal = {arXiv preprint arXiv:1802.01565},
month = {feb},
title = {{3D Snapshot Microscopy of Extended Objects}},
url = {http://arxiv.org/abs/1802.01565},
year = {2018}
}

Volumetric biological imaging often involves compromising high temporal resolution at the expense of high spatial resolution when popular scanning methods are used to capture 3D information. We introduce an integrated experimental and image reconstruction method for capturing dynamic 3D fluorescent extended objects as a series of synchronously measured 3D snapshots taken at the frame rate of the imaging camera. We employ multifocal microscopy (MFM) to simultaneously image at 25 focal planes and process this depth-encoded image to recover the 3D structure of extended objects, such as bacteria, using a sparsity-based reconstruction approach. The combined experimental and computational method produces image quality similar to confocal microscopy in a fraction of the acquisition time. In addition, our computational image reconstruction approach allows a simplified MFM optical design by correcting aberrations using the measured response to point sources. This "compressive" MFM acquisition and reconstruction method, where an image volume with roughly 8 million voxels is recovered from a single 1-megapixel captured image, enables straightforward study of dynamic processes in 3D, and as a simultaneous snapshot advances the state of the art in dynamic 3D microscopy.

@article{Michael2018,
abstract = {In this work we present a deep learning framework for video compressive sensing. The proposed formulation enables recovery of video frames in a few seconds at significantly improved reconstruction quality compared to previous approaches. Our investigation starts by learning a linear mapping between video sequences and corresponding measured frames which turns out to provide promising results. We then extend the linear formulation to deep fully-connected networks and explore the performance gains using deeper architectures. Our analysis is always driven by the applicability of the proposed framework on existing compressive video architectures. Extensive simulations on several video sequences document the superiority of our approach both quantitatively and qualitatively. Finally, our analysis offers insights into understanding how dataset sizes and number of layers affect reconstruction performance while raising a few points for future investigation.},
archivePrefix = {arXiv},
arxivId = {1603.04930},
author = {Iliadis, Michael and Spinoulas, Leonidas and Katsaggelos, Aggelos K.},
doi = {10.1016/j.dsp.2017.09.010},
eprint = {1603.04930},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Deep neural networks,Fully-connected networks,Video compressive sensing},
month = {jan},
pages = {9--18},
title = {{Deep fully-connected networks for video compressive sensing}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200417302130},
volume = {72},
year = {2018}
}

In this work we present a deep learning framework for video compressive sensing. The proposed formulation enables recovery of video frames in a few seconds at significantly improved reconstruction quality compared to previous approaches. Our investigation starts by learning a linear mapping between video sequences and corresponding measured frames which turns out to provide promising results. We then extend the linear formulation to deep fully-connected networks and explore the performance gains using deeper architectures. Our analysis is always driven by the applicability of the proposed framework on existing compressive video architectures. Extensive simulations on several video sequences document the superiority of our approach both quantitatively and qualitatively. Finally, our analysis offers insights into understanding how dataset sizes and number of layers affect reconstruction performance while raising a few points for future investigation.

@article{Neda2018a,
abstract = {Nonlinear unmixing of hyperspectral reflectance data is one of the key problems in quantitative imaging of painted works of art. The approach presented is to interrogate a hyperspectral image cube by first decomposing it into a set of reflectance curves representing pure basis pigments and second to estimate the scattering and absorption coefficients of each pigment in a given pixel to produce estimates of the component fractions. This two-step algorithm uses a deep neural network to qualitatively identify the constituent pigments in any unknown spectrum and, based on the pigment(s) present and Kubelka–Munk theory to estimate the pigment concentration on a per-pixel basis. Using hyperspectral data acquired on a set of mock-up paintings and a well-characterized illuminated folio from the 15th century, the performance of the proposed algorithm is demonstrated for pigment recognition and quantitative estimation of concentration.},
author = {Rohani, Neda and Pouyet, Emeline and Walton, Marc and Cossairt, Oliver and Katsaggelos, Aggelos K.},
doi = {10.1002/anie.201805135},
issn = {14337851},
journal = {Angewandte Chemie International Edition},
keywords = {deep neural network classification,heritage science,nonlinear unmixing Kubelka–Munk theory,visible hyperspectral imaging},
month = {aug},
number = {34},
pages = {10910--10914},
pmid = {29940088},
title = {{Nonlinear Unmixing of Hyperspectral Datasets for the Study of Painted Works of Art}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/anie.201805135},
volume = {57},
year = {2018}
}

Nonlinear unmixing of hyperspectral reflectance data is one of the key problems in quantitative imaging of painted works of art. The approach presented is to interrogate a hyperspectral image cube by first decomposing it into a set of reflectance curves representing pure basis pigments and second to estimate the scattering and absorption coefficients of each pigment in a given pixel to produce estimates of the component fractions. This two-step algorithm uses a deep neural network to qualitatively identify the constituent pigments in any unknown spectrum and, based on the pigment(s) present and Kubelka–Munk theory to estimate the pigment concentration on a per-pixel basis. Using hyperspectral data acquired on a set of mock-up paintings and a well-characterized illuminated folio from the 15th century, the performance of the proposed algorithm is demonstrated for pigment recognition and quantitative estimation of concentration.

@article{Emeline2018,
abstract = {Visible hyperspectral imaging (HSI) is a fast and non-invasive imaging method that has been adapted by the field of conservation science to study painted surfaces. By collecting reflectance spectra from a 2D surface, the resulting 3D hyperspectral data cube contains millions of recorded spectra. While processing such large amounts of spectra poses an analytical and computational challenge, it also opens new opportunities to apply powerful methods of multivariate analysis for data evaluation. With the intent of expanding current data treatment of hyperspectral datasets, an innovative approach for data reduction and visualization is presented in this article. It uses a statistical embedding method known as t-distributed stochastic neighbor embedding (t-SNE) to provide a non-linear representation of spectral features in a lower 2D space. The efficiency of the proposed method for painted surfaces from cultural heritage is established through the study of laboratory prepared paint mock-ups, and medieval French illuminated manuscript.},
author = {Pouyet, Emeline and Rohani, Neda and Katsaggelos, Aggelos K. and Cossairt, Oliver and Walton, Marc},
doi = {10.1515/pac-2017-0907},
issn = {1365-3075},
journal = {Pure and Applied Chemistry},
keywords = {ChemCultHerit,Data reduction and visualization,Illuminated manuscript,Multivariate analysis,T-distributed stochastic neighbor embedding,Visible hyperspectral imaging},
month = {feb},
number = {3},
pages = {493--506},
title = {{Innovative data reduction and visualization strategy for hyperspectral imaging datasets using t-SNE approach}},
url = {https://www.degruyter.com/document/doi/10.1515/pac-2017-0907/html},
volume = {90},
year = {2018}
}

Visible hyperspectral imaging (HSI) is a fast and non-invasive imaging method that has been adapted by the field of conservation science to study painted surfaces. By collecting reflectance spectra from a 2D surface, the resulting 3D hyperspectral data cube contains millions of recorded spectra. While processing such large amounts of spectra poses an analytical and computational challenge, it also opens new opportunities to apply powerful methods of multivariate analysis for data evaluation. With the intent of expanding current data treatment of hyperspectral datasets, an innovative approach for data reduction and visualization is presented in this article. It uses a statistical embedding method known as t-distributed stochastic neighbor embedding (t-SNE) to provide a non-linear representation of spectral features in a lower 2D space. The efficiency of the proposed method for painted surfaces from cultural heritage is established through the study of laboratory prepared paint mock-ups, and medieval French illuminated manuscript.

@article{Kuan2018,
abstract = {Despite recent advances, high performance single-shot 3D microscopy remains an elusive task. By introducing designed diffractive optical elements (DOEs), one is capable of converting a microscope into a 3D "kaleidoscope", in which case the snapshot image consists of an array of tiles and each tile focuses on different depths. However, the acquired multifocal microscopic (MFM) image suffers from multiple sources of degradation, which prevents MFM from further applications. We propose a unifying computational framework which simplifies the imaging system and achieves 3D reconstruction via computation. Our optical configuration omits chromatic correction grating and redesigns the multifocal grating to enlarge the tracking area. Our proposed setup features only one single grating in addition to a regular microscope. The aberration correction, along with Poisson and background denoising, are incorporated in our deconvolution-based fully-automated algorithm, which requires no empirical parameter-tuning. In experiments, we achieve the spatial resolutions of $0.35$um (lateral) and $0.5$um (axial), which are comparable to the resolution that can be achieved with confocal deconvolution microscopy. We demonstrate a 3D video of moving bacteria recorded at $25$ frames per second using our proposed computational multifocal microscopy technique.},
archivePrefix = {arXiv},
arxivId = {1809.01239},
author = {He, Kuan and Wang, Zihao and Huang, Xiang and Wang, Xiaolei and Yoo, Seunghwan and Ruiz, Pablo and Gdor, Itay and Selewa, Alan and Ferrier, Nicola J. and Scherer, Norbert and Hereld, Mark and Katsaggelos, Aggelos K. and Cossairt, Oliver},
doi = {10.1364/BOE.9.006477},
eprint = {1809.01239},
issn = {2156-7085},
journal = {Biomedical Optics Express},
month = {dec},
number = {12},
pages = {6477},
title = {{Computational multifocal microscopy}},
url = {https://opg.optica.org/abstract.cfm?URI=boe-9-12-6477},
volume = {9},
year = {2018}
}

Despite recent advances, high performance single-shot 3D microscopy remains an elusive task. By introducing designed diffractive optical elements (DOEs), one is capable of converting a microscope into a 3D "kaleidoscope", in which case the snapshot image consists of an array of tiles and each tile focuses on different depths. However, the acquired multifocal microscopic (MFM) image suffers from multiple sources of degradation, which prevents MFM from further applications. We propose a unifying computational framework which simplifies the imaging system and achieves 3D reconstruction via computation. Our optical configuration omits chromatic correction grating and redesigns the multifocal grating to enlarge the tracking area. Our proposed setup features only one single grating in addition to a regular microscope. The aberration correction, along with Poisson and background denoising, are incorporated in our deconvolution-based fully-automated algorithm, which requires no empirical parameter-tuning. In experiments, we achieve the spatial resolutions of $0.35$um (lateral) and $0.5$um (axial), which are comparable to the resolution that can be achieved with confocal deconvolution microscopy. We demonstrate a 3D video of moving bacteria recorded at $25$ frames per second using our proposed computational multifocal microscopy technique.

@article{Danielle2018,
abstract = {Microscopy and Microanalysis Volume 24 supplement s1: proceedings of microscopy & microanalysis 2018},
author = {Duggins, Danielle and Li, Fengqiang and Aalders, Maurice and Cossairt, Oliver and Katsaggelos, Aggelos and Walton, Marc},
doi = {10.1017/S1431927618011194},
issn = {1431-9276},
journal = {Microscopy and Microanalysis},
month = {aug},
number = {S1},
pages = {2142--2143},
title = {{A Novel OCT Design for Cultural Heritage Applications}},
url = {https://academic.oup.com/mam/article/24/S1/2142/6945740},
volume = {24},
year = {2018}
}

Microscopy and Microanalysis Volume 24 supplement s1: proceedings of microscopy & microanalysis 2018

@article{Alice2018,
abstract = {Traditionally, analytical methods have been used to solve imaging problems such as image restoration, inpainting, and superresolution (SR). In recent years, the fields of machine and deep learning have gained a lot of momentum in solving such imaging problems, often surpassing the performance provided by analytical approaches. Unlike analytical methods for which the problem is explicitly defined and domain-knowledge carefully engineered into the solution, deep neural networks (DNNs) do not benefit from such prior knowledge and instead make use of large data sets to learn the unknown solution to the inverse problem. In this article, we review deep-learning techniques for solving such inverse problems in imaging. More specifically, we review the popular neural network architectures used for imaging tasks, offering some insight as to how these deep-learning tools can solve the inverse problem. Furthermore, we address some fundamental questions, such as how deeplearning and analytical methods can be combined to provide better solutions to the inverse problem in addition to providing a discussion on the current limitations and future directions of the use of deep learning for solving inverse problem in imaging.},
author = {Lucas, Alice and Iliadis, Michael and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/MSP.2017.2760358},
issn = {1053-5888},
journal = {IEEE Signal Processing Magazine},
month = {jan},
number = {1},
pages = {20--36},
title = {{Using Deep Neural Networks for Inverse Problems in Imaging: Beyond Analytical Methods}},
url = {http://ieeexplore.ieee.org/document/8253590/},
volume = {35},
year = {2018}
}

Traditionally, analytical methods have been used to solve imaging problems such as image restoration, inpainting, and superresolution (SR). In recent years, the fields of machine and deep learning have gained a lot of momentum in solving such imaging problems, often surpassing the performance provided by analytical approaches. Unlike analytical methods for which the problem is explicitly defined and domain-knowledge carefully engineered into the solution, deep neural networks (DNNs) do not benefit from such prior knowledge and instead make use of large data sets to learn the unknown solution to the inverse problem. In this article, we review deep-learning techniques for solving such inverse problems in imaging. More specifically, we review the popular neural network architectures used for imaging tasks, offering some insight as to how these deep-learning tools can solve the inverse problem. Furthermore, we address some fundamental questions, such as how deeplearning and analytical methods can be combined to provide better solutions to the inverse problem in addition to providing a discussion on the current limitations and future directions of the use of deep learning for solving inverse problem in imaging.

@article{Itay2018,
abstract = {Accurate and rapid particle tracking is essential for addressing many research problems in single molecule and cellular biophysics and colloidal soft condensed matter physics. We developed a novel three-dimensional interferometric fluorescent particle tracking approach that does not require any sample scanning. By periodically shifting the interferometer phase, the information stored in the interference pattern of the emitted light allows localizing particles positions with nanometer resolution. This tracking protocol was demonstrated by measuring a known trajectory of a fluorescent bead with sub-5 nm axial localization error at 5 Hz. The interferometric microscopy was used to track the RecA protein in Bacillus subtilis bacteria to demonstrate its compatibility with biological systems.},
author = {Gdor, Itay and Wang, Xiaolei and Daddysman, Matthew and Yifat, Yuval and Wilton, Rosemarie and Hereld, Mark and Noirot-Gros, Marie-Fran{\c{c}}oise and Scherer, Norbert F.},
doi = {10.1364/OL.43.002819},
issn = {0146-9592},
journal = {Optics Letters},
month = {jun},
number = {12},
pages = {2819},
pmid = {29905697},
title = {{Particle tracking by repetitive phase-shift interferometric super resolution microscopy}},
url = {https://opg.optica.org/abstract.cfm?URI=ol-43-12-2819},
volume = {43},
year = {2018}
}

Accurate and rapid particle tracking is essential for addressing many research problems in single molecule and cellular biophysics and colloidal soft condensed matter physics. We developed a novel three-dimensional interferometric fluorescent particle tracking approach that does not require any sample scanning. By periodically shifting the interferometer phase, the information stored in the interference pattern of the emitted light allows localizing particles positions with nanometer resolution. This tracking protocol was demonstrated by measuring a known trajectory of a fluorescent bead with sub-5 nm axial localization error at 5 Hz. The interferometric microscopy was used to track the RecA protein in Bacillus subtilis bacteria to demonstrate its compatibility with biological systems.

@article{bahaadini2018machine,
abstract = {The detection of gravitational waves with ground-based laser-interferometric detectors requires sensitivity to changes in distance much smaller than the diameter of atomic nuclei. Though sophisticated machinery and techniques have been developed over the past few decades to isolate such instruments from non-astrophysical noise, the detectors are still susceptible to instrumental and environmental noise transients known as “glitches,” which hinder searches for transient gravitational waves. The Gravity Spy project is an effort to comprehensively classify the glitches that afflict gravitational wave detectors into morphological families by combining the strengths of machine learning algorithms and citizen scientists. This paper presents the initial Gravity Spy dataset used for citizen scientist and machine learning classification – a static, accessible, documented dataset for testing machine learning supervised classification. Previous versions of this dataset used in [8, 53] did not include all current classes and also for some of the classes, some samples were pruned and added. This set consists of time–frequency images of LIGO glitches and their associated metadata. These glitches are organized by time–frequency morphology into 22 classes for which descriptions and representative images are presented. Results from the application of state-of-the-art supervised classification methods to this dataset are presented in order to provide baselines for future glitch classification work. Standard splitting for training, validation, and testing sets are also presented to facilitate the comparison between different machine learning methods. The baseline methods are selected from both traditional and more recent deep learning approaches. An ensemble framework is developed that demonstrates that combining various classifiers can yield a more accurate model for classification. The ensemble classifier, trained with the standard training set, achieves 98.21% accuracy on the standard test set.},
author = {Bahaadini, S. and Noroozi, V. and Rohani, N. and Coughlin, S. and Zevin, M. and Smith, J.R. and Kalogera, V. and Katsaggelos, A.},
doi = {10.1016/j.ins.2018.02.068},
issn = {00200255},
journal = {Information Sciences},
keywords = {Classification,Dataset,Deep learning,Gravity Spy,Machine learning,aLIGO},
month = {may},
pages = {172--186},
publisher = {Elsevier},
title = {{Machine learning for Gravity Spy: Glitch classification and dataset}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0020025518301634},
volume = {444},
year = {2018}
}

The detection of gravitational waves with ground-based laser-interferometric detectors requires sensitivity to changes in distance much smaller than the diameter of atomic nuclei. Though sophisticated machinery and techniques have been developed over the past few decades to isolate such instruments from non-astrophysical noise, the detectors are still susceptible to instrumental and environmental noise transients known as “glitches,” which hinder searches for transient gravitational waves. The Gravity Spy project is an effort to comprehensively classify the glitches that afflict gravitational wave detectors into morphological families by combining the strengths of machine learning algorithms and citizen scientists. This paper presents the initial Gravity Spy dataset used for citizen scientist and machine learning classification – a static, accessible, documented dataset for testing machine learning supervised classification. Previous versions of this dataset used in [8, 53] did not include all current classes and also for some of the classes, some samples were pruned and added. This set consists of time–frequency images of LIGO glitches and their associated metadata. These glitches are organized by time–frequency morphology into 22 classes for which descriptions and representative images are presented. Results from the application of state-of-the-art supervised classification methods to this dataset are presented in order to provide baselines for future glitch classification work. Standard splitting for training, validation, and testing sets are also presented to facilitate the comparison between different machine learning methods. The baseline methods are selected from both traditional and more recent deep learning approaches. An ensemble framework is developed that demonstrates that combining various classifiers can yield a more accurate model for classification. The ensemble classifier, trained with the standard training set, achieves 98.21% accuracy on the standard test set.

@article{Kuan2018a,
abstract = {Realizing both high temporal and spatial resolution across a large volume is a key challenge for 3D fluorescent imaging. Towards achieving this objective, we introduce an interferometric multifocus microscopy (iMFM) system, a combination of multifocus microscopy (MFM) with two opposing objective lenses. We show that the proposed iMFM is capable of simultaneously producing multiple focal plane interferometry that provides axial super-resolution and hence isotropic 3D resolution with a single exposure. We design and simulate the iMFM microscope by employing two special diffractive optical elements. The point spread function of this new iMFM microscope is simulated and the image formation model is given. For reconstruction, we use the Richardson-Lucy deconvolution algorithm with total variation regularization for 3D extended object recovery, and a maximum likelihood estimator (MLE) for single molecule tracking. A method for determining an initial axial position of the molecule is also proposed to improve the convergence of the MLE. We demonstrate both theoretically and numerically that isotropic 3D nanoscopic localization accuracy is achievable with an axial imaging range of 2um when tracking a fluorescent molecule in three dimensions and that the diffraction limited axial resolution can be improved by 3–4 times in the single shot wide-field 3D extended object recovery. We believe that iMFM will be a useful tool in 3D dynamic event imaging that requires both high temporal and spatial resolution.},
author = {He, Kuan and Huang, Xiang and Wang, Xiaolei and Yoo, Seunghwan and Ruiz, Pablo and Gdor, Itay and Ferrier, Nicola J. and Scherer, Norbert and Hereld, Mark and Katsaggelos, Aggelos K. and Cossairt, Oliver},
doi = {10.1364/OE.26.027381},
issn = {1094-4087},
journal = {Optics Express},
month = {oct},
number = {21},
pages = {27381},
pmid = {30469808},
title = {{Design and simulation of a snapshot multi-focal interferometric microscope}},
url = {https://opg.optica.org/abstract.cfm?URI=oe-26-21-27381},
volume = {26},
year = {2018}
}

Realizing both high temporal and spatial resolution across a large volume is a key challenge for 3D fluorescent imaging. Towards achieving this objective, we introduce an interferometric multifocus microscopy (iMFM) system, a combination of multifocus microscopy (MFM) with two opposing objective lenses. We show that the proposed iMFM is capable of simultaneously producing multiple focal plane interferometry that provides axial super-resolution and hence isotropic 3D resolution with a single exposure. We design and simulate the iMFM microscope by employing two special diffractive optical elements. The point spread function of this new iMFM microscope is simulated and the image formation model is given. For reconstruction, we use the Richardson-Lucy deconvolution algorithm with total variation regularization for 3D extended object recovery, and a maximum likelihood estimator (MLE) for single molecule tracking. A method for determining an initial axial position of the molecule is also proposed to improve the convergence of the MLE. We demonstrate both theoretically and numerically that isotropic 3D nanoscopic localization accuracy is achievable with an axial imaging range of 2um when tracking a fluorescent molecule in three dimensions and that the diffraction limited axial resolution can be improved by 3–4 times in the single shot wide-field 3D extended object recovery. We believe that iMFM will be a useful tool in 3D dynamic event imaging that requires both high temporal and spatial resolution.

@article{AliBahrami2018,
abstract = {Aim An automatic resuscitation rhythm annotator (ARA) would facilitate and enhance retrospective analysis of resuscitation data, contributing to a better understanding of the interplay between therapy and patient response. The objective of this study was to define, implement, and demonstrate an ARA architecture for complete resuscitation episodes, including chest compression pauses (CC-pauses) and chest compression intervals (CC-intervals). Methods We analyzed 126.5 h of ECG and accelerometer-based chest-compression depth data from 281 out-of-hospital cardiac arrest (OHCA) patients. Data were annotated by expert reviewers into asystole (AS), pulseless electrical activity (PEA), pulse-generating rhythm (PR), ventricular fibrillation (VF), and ventricular tachycardia (VT). Clinical pulse annotations were based on patient-charts and impedance measurements. An ARA was developed for CC-pauses, and was used in combination with a chest compression artefact removal filter during CC-intervals. The performance of the ARA was assessed in terms of the unweighted mean of sensitivities (UMS). Results The UMS of the ARA were 75.0% during CC-pauses and 52.5% during CC-intervals, 55-points and 32.5-points over a random guess (20% for five categories). Filtering increased the UMS during CC-intervals by 5.2-points. Sensitivities for AS, PEA, PR, VF, and VT were 66.8%, 55.8%, 86.5%, 82.1% and 83.8% during CC-pauses; and 51.1%, 34.1%, 58.7%, 86.4%, and 32.1% during CC-intervals. Conclusions A general ARA architecture was defined and demonstrated on a comprehensive OHCA dataset. Results showed that semi-automatic resuscitation rhythm annotation, which may involve further revision/correction by clinicians for quality assurance, is feasible. The performance (UMS) dropped significantly during CC-intervals and sensitivity was lowest for PEA.},
author = {Rad, Ali Bahrami and Eftest{\o}l, Trygve and Irusta, Unai and Kval{\o}y, Jan Terje and Wik, Lars and Kramer-Johansen, Jo and Katsaggelos, Aggelos K. and Engan, Kjersti},
doi = {10.1016/j.resuscitation.2017.11.035},
issn = {03009572},
journal = {Resuscitation},
keywords = {Automatic resuscitation rhythm annotator,Cardiac arrest,Cardiac rhythm classification,Cardiopulmonary resuscitation},
month = {jan},
pages = {6--12},
pmid = {29122647},
title = {{An automatic system for the comprehensive retrospective analysis of cardiac rhythms in resuscitation episodes}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0300957217307177},
volume = {122},
year = {2018}
}

Aim An automatic resuscitation rhythm annotator (ARA) would facilitate and enhance retrospective analysis of resuscitation data, contributing to a better understanding of the interplay between therapy and patient response. The objective of this study was to define, implement, and demonstrate an ARA architecture for complete resuscitation episodes, including chest compression pauses (CC-pauses) and chest compression intervals (CC-intervals). Methods We analyzed 126.5 h of ECG and accelerometer-based chest-compression depth data from 281 out-of-hospital cardiac arrest (OHCA) patients. Data were annotated by expert reviewers into asystole (AS), pulseless electrical activity (PEA), pulse-generating rhythm (PR), ventricular fibrillation (VF), and ventricular tachycardia (VT). Clinical pulse annotations were based on patient-charts and impedance measurements. An ARA was developed for CC-pauses, and was used in combination with a chest compression artefact removal filter during CC-intervals. The performance of the ARA was assessed in terms of the unweighted mean of sensitivities (UMS). Results The UMS of the ARA were 75.0% during CC-pauses and 52.5% during CC-intervals, 55-points and 32.5-points over a random guess (20% for five categories). Filtering increased the UMS during CC-intervals by 5.2-points. Sensitivities for AS, PEA, PR, VF, and VT were 66.8%, 55.8%, 86.5%, 82.1% and 83.8% during CC-pauses; and 51.1%, 34.1%, 58.7%, 86.4%, and 32.1% during CC-intervals. Conclusions A general ARA architecture was defined and demonstrated on a comprehensive OHCA dataset. Results showed that semi-automatic resuscitation rhythm annotation, which may involve further revision/correction by clinicians for quality assurance, is feasible. The performance (UMS) dropped significantly during CC-intervals and sensitivity was lowest for PEA.

@article{Nima2017b,
abstract = {Artificial intelligence is revolutionizing our lives at an ever increasing pace. At the heart of this revolution is the recent advancements in deep neural networks (DNN), learning to perform sophisticated, high-level tasks. However, training DNNs requires massive amounts of data and is very computationally intensive. Gaining analytical understanding of the solutions found by DNNs can help us devise more efficient training algorithms, replacing the commonly used mthod of stochastic gradient descent (SGD). We analyze the dynamics of SGD and show that, indeed, direct computation of the solutions is possible in many cases. We show that a high performing setup used in DNNs introduces a separation of time-scales in the training dynamics, allowing SGD to train layers from the lowest (closest to input) to the highest. We then show that for each layer, the distribution of solutions found by SGD can be estimated using a class-based principal component analysis (PCA) of the layer's input. This finding allows us to forgo SGD entirely and directly derive the DNN parameters using this class-based PCA, which can be well estimated using significantly less data than SGD. We implement these results on image datasets MNIST, CIFAR10 and CIFAR100 and find that, in fact, layers derived using our class-based PCA perform comparable or superior to neural networks of the same size and architecture trained using SGD. We also confirm that the class-based PCA often converges using a fraction of the data required for SGD. Thus, using our method training time can be reduced both by requiring less training data than SGD, and by eliminating layers in the costly backpropagation step of the training.},
archivePrefix = {arXiv},
arxivId = {1703.04757},
author = {Dehmamy, Nima and Rohani, Neda and Katsaggelos, Aggelos and Nima, Dehmamy and Neda, Rohani and Aggelos, Katsaggelos},
eprint = {1703.04757},
journal = {arXiv preprint arXiv:1703.04757},
keywords = {neural networks},
month = {mar},
title = {{Separation of time scales and direct computation of weights in deep neural networks}},
url = {http://arxiv.org/abs/1703.04757},
volume = {1703},
year = {2017}
}

Artificial intelligence is revolutionizing our lives at an ever increasing pace. At the heart of this revolution is the recent advancements in deep neural networks (DNN), learning to perform sophisticated, high-level tasks. However, training DNNs requires massive amounts of data and is very computationally intensive. Gaining analytical understanding of the solutions found by DNNs can help us devise more efficient training algorithms, replacing the commonly used mthod of stochastic gradient descent (SGD). We analyze the dynamics of SGD and show that, indeed, direct computation of the solutions is possible in many cases. We show that a high performing setup used in DNNs introduces a separation of time-scales in the training dynamics, allowing SGD to train layers from the lowest (closest to input) to the highest. We then show that for each layer, the distribution of solutions found by SGD can be estimated using a class-based principal component analysis (PCA) of the layer's input. This finding allows us to forgo SGD entirely and directly derive the DNN parameters using this class-based PCA, which can be well estimated using significantly less data than SGD. We implement these results on image datasets MNIST, CIFAR10 and CIFAR100 and find that, in fact, layers derived using our class-based PCA perform comparable or superior to neural networks of the same size and architecture trained using SGD. We also confirm that the class-based PCA often converges using a fraction of the data required for SGD. Thus, using our method training time can be reduced both by requiring less training data than SGD, and by eliminating layers in the costly backpropagation step of the training.

@article{AliBahrami2017,
abstract = {Objective: There is a need to monitor the heart rhythm in resuscitation to improve treatment quality. Resuscitation rhythms are categorized into: ventricular tachycardia (VT), ventricular fibrillation (VF), pulseless electrical activity (PEA), asystole (AS), and pulse-generating rhythm (PR). Manual annotation of rhythms is time-consuming and infeasible for large datasets. Our objective was to develop ECG-based algorithms for the retrospective and automatic classification of resuscitation cardiac rhythms. Methods: The dataset consisted of 1631 3-s ECG segments with clinical rhythm annotations, obtained from 298 out-of-hospital cardiac arrest patients. In total, 47 wavelet- and time-domain-based features were computed from the ECG. Features were selected using a wrapper-based feature selection architecture. Classifiers based on Bayesian decision theory, k-nearest neighbor, k-local hyperplane distance nearest neighbor, artificial neural network (ANN), and ensemble of decision trees were studied. Results: The best results were obtained for ANN classifier with Bayesian regularization backpropagation training algorithm with 14 features, which forms the proposed algorithm. The overall accuracy for the proposed algorithm was 78.5%. The sensitivities (and positive-predictive-values) for AS, PEA, PR, VF, and VT were 88.7% (91.0%), 68.9% (70.4%), 65.9% (69.0%), 86.2% (83.8%), and 78.8% (72.9%), respectively. Conclusions: The results demonstrate that it is possible to classify resuscitation cardiac rhythms automatically, but the accuracy for the organized rhythms (PEA and PR) is low. Significance: We have made an important step toward making classification of resuscitation rhythms more efficient in the sense of minimal feedback from human experts.},
author = {Rad, Ali Bahrami and Eftestol, Trygve and Engan, Kjersti and Irusta, Unai and Kvaloy, Jan Terje and Kramer-Johansen, Jo and Wik, Lars and Katsaggelos, Aggelos K.},
doi = {10.1109/TBME.2017.2688380},
issn = {0018-9294},
journal = {IEEE Transactions on Biomedical Engineering},
keywords = {Cardiac arrest,cardiac rhythm classification,cardiopulmonary resuscitation,feature extraction/selection,nested cross-validation},
month = {oct},
number = {10},
pages = {2411--2418},
pmid = {28371771},
title = {{ECG-Based Classification of Resuscitation Cardiac Rhythms for Retrospective Data Analysis}},
url = {http://ieeexplore.ieee.org/document/7890478/},
volume = {64},
year = {2017}
}

Objective: There is a need to monitor the heart rhythm in resuscitation to improve treatment quality. Resuscitation rhythms are categorized into: ventricular tachycardia (VT), ventricular fibrillation (VF), pulseless electrical activity (PEA), asystole (AS), and pulse-generating rhythm (PR). Manual annotation of rhythms is time-consuming and infeasible for large datasets. Our objective was to develop ECG-based algorithms for the retrospective and automatic classification of resuscitation cardiac rhythms. Methods: The dataset consisted of 1631 3-s ECG segments with clinical rhythm annotations, obtained from 298 out-of-hospital cardiac arrest patients. In total, 47 wavelet- and time-domain-based features were computed from the ECG. Features were selected using a wrapper-based feature selection architecture. Classifiers based on Bayesian decision theory, k-nearest neighbor, k-local hyperplane distance nearest neighbor, artificial neural network (ANN), and ensemble of decision trees were studied. Results: The best results were obtained for ANN classifier with Bayesian regularization backpropagation training algorithm with 14 features, which forms the proposed algorithm. The overall accuracy for the proposed algorithm was 78.5%. The sensitivities (and positive-predictive-values) for AS, PEA, PR, VF, and VT were 88.7% (91.0%), 68.9% (70.4%), 65.9% (69.0%), 86.2% (83.8%), and 78.8% (72.9%), respectively. Conclusions: The results demonstrate that it is possible to classify resuscitation cardiac rhythms automatically, but the accuracy for the organized rhythms (PEA and PR) is low. Significance: We have made an important step toward making classification of resuscitation rhythms more efficient in the sense of minimal feedback from human experts.

@article{pouyet2017revealing,
abstract = {Reading the content of hidden texts from ancient manuscripts has become an increasingly important endeavor thanks to the variety of non-destructive analytical tools and image processing routines available for this task. In this study, portable macro X-Ray Fluorescence (MA-XRF-tube), Visible Hyperspectral Imaging (HSI) together with Synchrotron based macro X-Ray Fluorescence (MA-XRF-SR) are combined with signal processing methods to reveal the biography of a degraded manuscript recycled as binding material for a 16th century printed edition of Hesiod's Works and Days. The analytical techniques allow visualizing the hidden text, revealing passages from the Institutes Justinian, a 6th century A.D codification of the Roman Law, with further marginal comments on medieval Canon Law. In addition, the identification of the materials (e.g. pigments, inks) part of the original manuscript together with their sequence of use are revealed: i) the preparation of the parchment using a Ca-based preparation layer, ii) drawing of ruled guide lines, using a Pb-based pen or ink, iii) writing of the main text using a rich Fe-gall ink with modulating color pigments (Hg-, Cu- and Pb- based) and iv) addition of two types of comments to the main text, one of the ink used for the comments being a Fe-gall ink rich in Cu.},
author = {Pouyet, E. and Devine, S. and Grafakos, T. and Kieckhefer, R. and Salvant, J. and Smieska, L. and Woll, A. and Katsaggelos, A. and Cossairt, O. and Walton, M.},
doi = {10.1016/j.aca.2017.06.016},
issn = {00032670},
journal = {Analytica Chimica Acta},
keywords = {Data fusion,Hidden manuscript,Hyperspectral,Inverse learning,Synchrotron,X-Ray Fluorescence},
month = {aug},
pages = {20--30},
pmid = {28734360},
publisher = {Elsevier},
title = {{Revealing the biography of a hidden medieval manuscript using synchrotron and conventional imaging techniques}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0003267017307201},
volume = {982},
year = {2017}
}

Reading the content of hidden texts from ancient manuscripts has become an increasingly important endeavor thanks to the variety of non-destructive analytical tools and image processing routines available for this task. In this study, portable macro X-Ray Fluorescence (MA-XRF-tube), Visible Hyperspectral Imaging (HSI) together with Synchrotron based macro X-Ray Fluorescence (MA-XRF-SR) are combined with signal processing methods to reveal the biography of a degraded manuscript recycled as binding material for a 16th century printed edition of Hesiod's Works and Days. The analytical techniques allow visualizing the hidden text, revealing passages from the Institutes Justinian, a 6th century A.D codification of the Roman Law, with further marginal comments on medieval Canon Law. In addition, the identification of the materials (e.g. pigments, inks) part of the original manuscript together with their sequence of use are revealed: i) the preparation of the parchment using a Ca-based preparation layer, ii) drawing of ruled guide lines, using a Pb-based pen or ink, iii) writing of the main text using a rich Fe-gall ink with modulating color pigments (Hg-, Cu- and Pb- based) and iv) addition of two types of comments to the main text, one of the ink used for the comments being a Fe-gall ink rich in Cu.

@article{Matthew2017a,
abstract = {The image processing technique known as super-resolution (SR), which attempts to increase the effective pixel sampling density of a digital imager, has gained rapid popularity over the last decade. The majority of literature focuses on its ability to provide results that are visually pleasing to a human observer. In this paper, we instead examine the ability of SR to improve the resolution-critical capability of an imaging system to perform a classification task from a remote location, specifically from an airborne camera. In order to focus the scope of the study, we address and quantify results for the narrow case of text classification. However, we expect the results generalize to a large set of related, remote classification tasks. We generate theoretical results through simulation, which are corroborated by experiments with a camera mounted on a DJI Phantom 3 quadcopter.},
author = {Woods, Matthew and Katsaggelos, Aggelos},
doi = {10.1364/JOSAA.34.000203},
issn = {1084-7529},
journal = {Journal of the Optical Society of America A},
month = {feb},
number = {2},
pages = {203},
pmid = {28157846},
title = {{Remote classification from an airborne camera using image super-resolution}},
url = {https://opg.optica.org/abstract.cfm?URI=josaa-34-2-203},
volume = {34},
year = {2017}
}

The image processing technique known as super-resolution (SR), which attempts to increase the effective pixel sampling density of a digital imager, has gained rapid popularity over the last decade. The majority of literature focuses on its ability to provide results that are visually pleasing to a human observer. In this paper, we instead examine the ability of SR to improve the resolution-critical capability of an imaging system to perform a classification task from a remote location, specifically from an airborne camera. In order to focus the scope of the study, we address and quantify results for the narrow case of text classification. However, we expect the results generalize to a large set of related, remote classification tasks. We generate theoretical results through simulation, which are corroborated by experiments with a camera mounted on a DJI Phantom 3 quadcopter.

@article{Huyen2017,
abstract = {Objectives Newborn deaths are reported to be caused mainly by birth asphyxia. Information learned from ventilation and other treatment could help increase survival rate of newborns in need of resuscitation. Characteristics of manual bag-mask ventilation have been studied in our previous works. However, other resuscitation activities could have important impacts as well. This paper illustrates the classification of several predefined resuscitation activities using information from acceleration and ECG signal. Methods Time and frequency domain features were extracted from the acceleration and ECG signals. A 2-stage classifier was trained on data of manually annotated activities by observing videos of 30 resuscitation babies. Leave-one-out cross validation was used: for each fold, the classifier was trained on activities of 29 patients and tested on activities of 1 patient. Results The average accuracy of the classification of activities is 79%. Conclusions The performance of the classification algorithms indicates that it is possible to use ECG and acceleration signals to automatically derive useful information regarding resuscitation activities.},
author = {Vu, Huyen and Engan, Kjersti and Eftest{\o}l, Trygve and Katsaggelos, Aggelos and Jatosh, Samwel and Kusulla, Simeon and Mduma, Estomih and Kidanto, Hussein and Ersdal, Hege},
doi = {10.1016/j.bspc.2017.03.004},
issn = {17468094},
journal = {Biomedical Signal Processing and Control},
keywords = {Acceleration signal,Birth asphyxia,Decision tree,ECG signal,Linear discriminant analysis,Newborns,Short time energy,Wavelet packet decomposition},
month = {jul},
pages = {20--26},
title = {{Automatic classification of resuscitation activities on birth-asphyxiated newborns using acceleration and ECG signals}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S174680941730054X},
volume = {36},
year = {2017}
}

Objectives Newborn deaths are reported to be caused mainly by birth asphyxia. Information learned from ventilation and other treatment could help increase survival rate of newborns in need of resuscitation. Characteristics of manual bag-mask ventilation have been studied in our previous works. However, other resuscitation activities could have important impacts as well. This paper illustrates the classification of several predefined resuscitation activities using information from acceleration and ECG signal. Methods Time and frequency domain features were extracted from the acceleration and ECG signals. A 2-stage classifier was trained on data of manually annotated activities by observing videos of 30 resuscitation babies. Leave-one-out cross validation was used: for each fold, the classifier was trained on activities of 29 patients and tested on activities of 1 patient. Results The average accuracy of the classification of activities is 79%. Conclusions The performance of the classification algorithms indicates that it is possible to use ECG and acceleration signals to automatically derive useful information regarding resuscitation activities.

@article{Xu2017,
abstract = {Expectation Maximization (EM) based inference has already proven to be a very powerful tool to solve blind image deconvolution (BID) problems. Unfortunately, three important problems still impede the application of EM in BID: the undesirable saddle points and local minima caused by highly nonconvex priors, the instability around zero of some of the most interesting sparsity promoting priors, and the intrinsic high computational cost of the corresponding BID algorithm. In this paper we first show how Super Gaussian priors can be made numerically tractable around zero by introducing the family of Huber Super Gaussian priors and then present a fast EM based blind deconvolution method formulated in the image space. In the proposed computational approach, image and kernel estimation are performed by using the Alternating Direction Method of Multipliers (ADMM), which allows to exploit the advantages of FFT computation. For highly nonconvex priors, we propose a Smooth ADMM (SADMM) approach to avoid poor BID estimates. Extensive experiments demonstrate that the proposed method significantly outperforms state-of-the-art BID methods in terms of quality of the reconstructions and speed.},
author = {Zhou, Xu and Vega, Miguel and Zhou, Fugen and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.dsp.2016.08.008},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Blind deconvolution,Image deblurring,Image restoration,Variational Bayesian},
month = {jan},
pages = {122--133},
title = {{Fast Bayesian blind deconvolution with Huber Super Gaussian priors}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200416301245},
volume = {60},
year = {2017}
}

Expectation Maximization (EM) based inference has already proven to be a very powerful tool to solve blind image deconvolution (BID) problems. Unfortunately, three important problems still impede the application of EM in BID: the undesirable saddle points and local minima caused by highly nonconvex priors, the instability around zero of some of the most interesting sparsity promoting priors, and the intrinsic high computational cost of the corresponding BID algorithm. In this paper we first show how Super Gaussian priors can be made numerically tractable around zero by introducing the family of Huber Super Gaussian priors and then present a fast EM based blind deconvolution method formulated in the image space. In the proposed computational approach, image and kernel estimation are performed by using the Alternating Direction Method of Multipliers (ADMM), which allows to exploit the advantages of FFT computation. For highly nonconvex priors, we propose a Smooth ADMM (SADMM) approach to avoid poor BID estimates. Extensive experiments demonstrate that the proposed method significantly outperforms state-of-the-art BID methods in terms of quality of the reconstructions and speed.

@article{kemner2017integrated,
author = {Kemner, K. M. and Hereld, M. and Scherer, N. and Selewa, A. and Wang, X. and Gdor, I. and Daddysman, M. and Jureller, J. and Huynh, T. and Cossairt, O. and Katsaggelos, A. and He, K. and Yoo, S. and Matsuda, N. and Glick, B. and Riviere, P. La and Austin, J. and Day, K. and Chandler, T. and Papanikou, S. and Ferrier, N. and Sholto-Douglas, D. and Gursoy, D. and Antipova, O. and Soriano, C. and O'Brien, S. and Wilton, R. and Ahrendt, A. and Asplund, M. and Zerbs, S. and Noirot, P. and Atkins, C. and Babnigg, G. and Johnson, J. and Shinde, S. and Korajczyk, P. and Noirot, M. F.},
doi = {10.1017/S1431927617002380},
issn = {1431-9276},
journal = {Microscopy and Microanalysis},
month = {jul},
number = {S1},
pages = {340--341},
publisher = {Cambridge University Press},
title = {{Integrated Dynamic 3D Imaging of Microbial Processes and Communities in Rhizosphere Environments: The Argonne Small Worlds Project}},
url = {https://academic.oup.com/mam/article/23/S1/340/6944835},
volume = {23},
year = {2017}
}

@article{Marc2014,
author = {Walton, Marc and Pastorelli, Gianluca and Pouyet, Emeline and Katsaggelos, Aggelos and Cossairt, Oliver and Dai, Qiqin},
doi = {10.1107/S2053273317089823},
issn = {2053-2733},
journal = {Acta Crystallographica Section A Foundations and Advances},
month = {dec},
number = {a2},
pages = {C591--C591},
title = {{Getting more for less: adaptive X-ray fluorescence sampling for imaging}},
url = {http://scripts.iucr.org/cgi-bin/paper?S2053273317089823},
volume = {73},
year = {2017}
}

@article{Reza2017,
abstract = {Sparse coding of images is traditionally done by cutting them into small patches and representing each patch individually over some dictionary given a pre-determined number of nonzero coefficients to use for each patch. In lack of a way to effectively distribute a total number (or global budget) of nonzero coefficients across all patches, current sparse recovery algorithms distribute the global budget equally across all patches despite the wide range of differences in structural complexity among them. In this work we propose a new framework for joint sparse representation and recovery of all image patches simultaneously. We also present two novel global hard thresholding algorithms, based on the notion of variable splitting, for solving the joint sparse model. Experimentation using both synthetic and real data shows effectiveness of the proposed framework for sparse image representation and denoising tasks. Additionally, time complexity analysis of the proposed algorithms indicate high scalability of both algorithms, making them favorable to use on large megapixel images.},
archivePrefix = {arXiv},
arxivId = {1705.09816},
author = {Borhani, Reza and Watt, Jeremy and Katsaggelos, Aggelos},
eprint = {1705.09816},
journal = {arXiv preprint arXiv:1705.09816},
month = {may},
title = {{Global hard thresholding algorithms for joint sparse image representation and denoising}},
url = {http://arxiv.org/abs/1705.09816},
year = {2017}
}

Sparse coding of images is traditionally done by cutting them into small patches and representing each patch individually over some dictionary given a pre-determined number of nonzero coefficients to use for each patch. In lack of a way to effectively distribute a total number (or global budget) of nonzero coefficients across all patches, current sparse recovery algorithms distribute the global budget equally across all patches despite the wide range of differences in structural complexity among them. In this work we propose a new framework for joint sparse representation and recovery of all image patches simultaneously. We also present two novel global hard thresholding algorithms, based on the notion of variable splitting, for solving the joint sparse model. Experimentation using both synthetic and real data shows effectiveness of the proposed framework for sparse image representation and denoising tasks. Additionally, time complexity analysis of the proposed algorithms indicate high scalability of both algorithms, making them favorable to use on large megapixel images.

@article{Qiqin2016,
abstract = {In this paper, we propose two multiple-frame super-resolution (SR) algorithms based on dictionary learning (DL) and motion estimation. First, we adopt the use of video bilevel DL, which has been used for single-frame SR. It is extended to multiple frames by using motion estimation with sub-pixel accuracy. We propose a batch and a temporally recursive multi-frame SR algorithm, which improves over single-frame SR. Finally, we propose a novel DL algorithm utilizing consecutive video frames, rather than still images or individual video frames, which further improves the performance of the video SR algorithms. Extensive experimental comparisons with the state-of-the-art SR algorithms verify the effectiveness of our proposed multiple-frame video SR approach.},
author = {Dai, Qiqin and Yoo, Seunghwan and Kappeler, Armin and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2016.2631339},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Video super-resolution,dictionary learning,motion estimation,optical flow,sparse coding},
month = {feb},
number = {2},
pages = {765--781},
pmid = {27893388},
title = {{Sparse Representation-Based Multiple Frame Video Super-Resolution}},
url = {http://ieeexplore.ieee.org/document/7752984/},
volume = {26},
year = {2017}
}

In this paper, we propose two multiple-frame super-resolution (SR) algorithms based on dictionary learning (DL) and motion estimation. First, we adopt the use of video bilevel DL, which has been used for single-frame SR. It is extended to multiple frames by using motion estimation with sub-pixel accuracy. We propose a batch and a temporally recursive multi-frame SR algorithm, which improves over single-frame SR. Finally, we propose a novel DL algorithm utilizing consecutive video frames, rather than still images or individual video frames, which further improves the performance of the video SR algorithms. Extensive experimental comparisons with the state-of-the-art SR algorithms verify the effectiveness of our proposed multiple-frame video SR approach.

@article{Michael2017a,
abstract = {In this paper, we propose an iterative method to address the face identification problem with block occlusions. Our approach utilizes a robust representation based on two characteristics in order to model contiguous errors (e.g., block occlusion) effectively. The first fits to the errors a distribution described by a tailored loss function. The second describes the error image as having a specific structure (resulting in low-rank in comparison with image size). We will show that this joint characterization is effective for describing errors with spatial continuity. Our approach is computationally efficient due to the utilization of the alternating direction method of multipliers. A special case of our fast iterative algorithm leads to the robust representation method, which is normally used to handle non-contiguous errors (e.g., pixel corruption). Extensive results on representative face databases (in constrained and unconstrained environments) document the effectiveness of our method over existing robust representation methods with respect to both identification rates and computational time.},
archivePrefix = {arXiv},
arxivId = {1605.02266},
author = {Iliadis, Michael and Wang, Haohong and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2017.2675206},
eprint = {1605.02266},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Face identification,Iterative reweighted coding,Low-rank estimation,Robust representation},
month = {may},
number = {5},
pages = {2203--2218},
pmid = {28252401},
title = {{Robust and Low-Rank Representation for Fast Face Identification With Occlusions}},
url = {http://ieeexplore.ieee.org/document/7864430/},
volume = {26},
year = {2017}
}

In this paper, we propose an iterative method to address the face identification problem with block occlusions. Our approach utilizes a robust representation based on two characteristics in order to model contiguous errors (e.g., block occlusion) effectively. The first fits to the errors a distribution described by a tailored loss function. The second describes the error image as having a specific structure (resulting in low-rank in comparison with image size). We will show that this joint characterization is effective for describing errors with spatial continuity. Our approach is computationally efficient due to the utilization of the alternating direction method of multipliers. A special case of our fast iterative algorithm leads to the robust representation method, which is normally used to handle non-contiguous errors (e.g., pixel corruption). Extensive results on representative face databases (in constrained and unconstrained environments) document the effectiveness of our method over existing robust representation methods with respect to both identification rates and computational time.

@article{Qiqin2017,
abstract = {X-Ray fluorescence (XRF) scanning of works of art is becoming an increasing popular non-destructive analytical method. The high quality XRF spectra is necessary to obtain significant information on both major and minor elements used for characterization and provenance analysis. However, there is a trade-off between the spatial resolution of an XRF scan and the Signal-to-Noise Ratio (SNR) of each pixel's spectrum, due to the limited scanning time. In this project, we propose an XRF image super-resolution method to address this trade-off, thus obtaining a high spatial resolution XRF scan with high SNR. We fuse a low resolution XRF image and a conventional RGB highresolution image into a product of both high spatial and high spectral resolution XRF image. There is no guarantee of a one to one mapping between XRF spectrum and RGB color since, for instance, paintings with hidden layers cannot be detected in visible but can in X-ray wavelengths. We separate the XRF image into the visible and non-visible components. The spatial resolution of the visible component is increased utilizing the high-resolution RGB image while the spatial resolution of the non-visible component is increased using a total variation superresolution method. Finally, the visible and non-visible components are combined to obtain the final result.},
author = {Dai, Qiqin and Pouyet, Emeline and Cossairt, Oliver and Walton, Marc and Katsaggelos, Aggelos K.},
doi = {10.1109/TCI.2017.2703987},
issn = {2333-9403},
journal = {IEEE Transactions on Computational Imaging},
month = {sep},
number = {3},
pages = {432--444},
title = {{Spatial-Spectral Representation for X-Ray Fluorescence Image Super-Resolution}},
url = {http://ieeexplore.ieee.org/document/7927468/},
volume = {3},
year = {2017}
}

X-Ray fluorescence (XRF) scanning of works of art is becoming an increasing popular non-destructive analytical method. The high quality XRF spectra is necessary to obtain significant information on both major and minor elements used for characterization and provenance analysis. However, there is a trade-off between the spatial resolution of an XRF scan and the Signal-to-Noise Ratio (SNR) of each pixel's spectrum, due to the limited scanning time. In this project, we propose an XRF image super-resolution method to address this trade-off, thus obtaining a high spatial resolution XRF scan with high SNR. We fuse a low resolution XRF image and a conventional RGB highresolution image into a product of both high spatial and high spectral resolution XRF image. There is no guarantee of a one to one mapping between XRF spectrum and RGB color since, for instance, paintings with hidden layers cannot be detected in visible but can in X-ray wavelengths. We separate the XRF image into the visible and non-visible components. The spatial resolution of the visible component is increased utilizing the high-resolution RGB image while the spatial resolution of the non-visible component is increased using a total variation superresolution method. Finally, the visible and non-visible components are combined to obtain the final result.

@article{Juan2017a,
abstract = {Recent work in signal processing in general and image processing in particular deals with sparse representation related problems. Two such problems are of paramount importance: an overriding need for designing a well-suited overcomplete dictionary containing a redundant set of atoms-i.e., basis signals-and how to find a sparse representation of a given signal with respect to the chosen dictionary. Dictionary learning techniques, among which we find the popular K-singular value decomposition algorithm, tackle these problems by adapting a dictionary to a set of training data. A common drawback of such techniques is the need for parameter-tuning. In order to overcome this limitation, we propose a fullyautomated Bayesian method that considers the uncertainty of the estimates and produces a sparse representation of the data without prior information on the number of non-zeros in each representation vector. We follow a Bayesian approach that uses a three-tiered hierarchical prior to enforce sparsity on the representations and develop an efficient variational inference framework that reduces computational complexity. Furthermore, we describe a greedy approach that speeds up the whole process. Finally, we present experimental results that show superior performance on two different applications with real images: denoising and inpainting.},
author = {Serra, Juan G. and Testa, Matteo and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2017.2681436},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian modeling,Denoising,Dictionary learning,Inpainting,Sparse representation,Variational inference,k-SVD},
month = {jul},
number = {7},
pages = {3344--3359},
pmid = {28362587},
title = {{Bayesian K-SVD Using Fast Variational Inference}},
url = {http://ieeexplore.ieee.org/document/7875464/},
volume = {26},
year = {2017}
}

Recent work in signal processing in general and image processing in particular deals with sparse representation related problems. Two such problems are of paramount importance: an overriding need for designing a well-suited overcomplete dictionary containing a redundant set of atoms-i.e., basis signals-and how to find a sparse representation of a given signal with respect to the chosen dictionary. Dictionary learning techniques, among which we find the popular K-singular value decomposition algorithm, tackle these problems by adapting a dictionary to a set of training data. A common drawback of such techniques is the need for parameter-tuning. In order to overcome this limitation, we propose a fullyautomated Bayesian method that considers the uncertainty of the estimates and produces a sparse representation of the data without prior information on the number of non-zeros in each representation vector. We follow a Bayesian approach that uses a three-tiered hierarchical prior to enforce sparsity on the representations and develop an efficient variational inference framework that reduces computational complexity. Furthermore, we describe a greedy approach that speeds up the whole process. Finally, we present experimental results that show superior performance on two different applications with real images: denoising and inpainting.

@article{Gursoy2017,
abstract = {As x-ray and electron tomography is pushed further into the nanoscale, the limitations of rotation stages become more apparent, leading to challenges in the alignment of the acquired projection images. Here we present an approach for rapid post-acquisition alignment of these projections to obtain high quality three-dimensional images. Our approach is based on a joint estimation of alignment errors, and the object, using an iterative refinement procedure. With simulated data where we know the alignment error of each projection image, our approach shows a residual alignment error that is a factor of a thousand smaller, and it reaches the same error level in the reconstructed image in less than half the number of iterations. We then show its application to experimental data in x-ray and electron nanotomography.},
author = {Gursoy, Doga and Hong, Young P. and He, Kuan and Hujsak, Karl and Yoo, Seunghwan and Chen, Si and Li, Yue and Ge, Mingyuan and Miller, Lisa M. and Chu, Yong S. and {De Andrade}, Vincent and He, Kai and Cossairt, Oliver and Katsaggelos, Aggelos K. and Jacobsen, Chris},
doi = {10.1038/s41598-017-12141-9},
issn = {2045-2322},
journal = {Scientific Reports},
month = {sep},
number = {1},
pages = {11818},
pmid = {28924196},
title = {{Rapid alignment of nanotomography data using joint iterative reconstruction and reprojection}},
url = {https://www.nature.com/articles/s41598-017-12141-9},
volume = {7},
year = {2017}
}

As x-ray and electron tomography is pushed further into the nanoscale, the limitations of rotation stages become more apparent, leading to challenges in the alignment of the acquired projection images. Here we present an approach for rapid post-acquisition alignment of these projections to obtain high quality three-dimensional images. Our approach is based on a joint estimation of alignment errors, and the object, using an iterative refinement procedure. With simulated data where we know the alignment error of each projection image, our approach shows a residual alignment error that is a factor of a thousand smaller, and it reaches the same error level in the reconstructed image in less than half the number of iterations. We then show its application to experimental data in x-ray and electron nanotomography.

@article{Michael2017,
abstract = {With the first direct detection of gravitational waves, the advanced laser interferometer gravitational-wave observatory (LIGO) has initiated a new field of astronomy by providing an alternative means of sensing the universe. The extreme sensitivity required to make such detections is achieved through exquisite isolation of all sensitive components of LIGO from non-gravitational-wave disturbances. Nonetheless, LIGO is still susceptible to a variety of instrumental and environmental sources of noise that contaminate the data. Of particular concern are noise features known as glitches, which are transient and non-Gaussian in their nature, and occur at a high enough rate so that accidental coincidence between the two LIGO detectors is non-negligible. Glitches come in a wide range of time-frequency-amplitude morphologies, with new morphologies appearing as the detector evolves. Since they can obscure or mimic true gravitational-wave signals, a robust characterization of glitches is paramount in the effort to achieve the gravitational-wave detection rates that are predicted by the design sensitivity of LIGO. This proves a daunting task for members of the LIGO Scientific Collaboration alone due to the sheer amount of data. In this paper we describe an innovative project that combines crowdsourcing with machine learning to aid in the challenging task of categorizing all of the glitches recorded by the LIGO detectors. Through the Zooniverse platform, we engage and recruit volunteers from the public to categorize images of time-frequency representations of glitches into pre-identified morphological classes and to discover new classes that appear as the detectors evolve. In addition, machine learning algorithms are used to categorize images after being trained on human-classified examples of the morphological classes. Leveraging the strengths of both classification methods, we create a combined method with the aim of improving the efficiency and accuracy of each individual classifier. The resulting classification and characterization should help LIGO scientists to identify causes of glitches and subsequently eliminate them from the data or the detector entirely, thereby improving the rate and accuracy of gravitational-wave observations. We demonstrate these methods using a small subset of data from LIGO's first observing run.},
archivePrefix = {arXiv},
arxivId = {1611.04596},
author = {Zevin, M. and Coughlin, S. and Bahaadini, S. and Besler, E. and Rohani, N. and Allen, S. and Cabero, M. and Crowston, K. and Katsaggelos, A. K. and Larson, S. L. and Lee, T. K. and Lintott, C. and Littenberg, T. B. and Lundgren, A. and {\O}sterlund, C and Smith, J. R. and Trouille, L. and Kalogera, V.},
doi = {10.1088/1361-6382/aa5cea},
eprint = {1611.04596},
issn = {0264-9381},
journal = {Classical and Quantum Gravity},
keywords = {LIGO,citizen science,detector characterization,gravitational waves,machine learning},
month = {mar},
number = {6},
pages = {064003},
title = {{Gravity Spy: integrating advanced LIGO detector characterization, machine learning, and citizen science}},
url = {https://iopscience.iop.org/article/10.1088/1361-6382/aa5cea},
volume = {34},
year = {2017}
}

With the first direct detection of gravitational waves, the advanced laser interferometer gravitational-wave observatory (LIGO) has initiated a new field of astronomy by providing an alternative means of sensing the universe. The extreme sensitivity required to make such detections is achieved through exquisite isolation of all sensitive components of LIGO from non-gravitational-wave disturbances. Nonetheless, LIGO is still susceptible to a variety of instrumental and environmental sources of noise that contaminate the data. Of particular concern are noise features known as glitches, which are transient and non-Gaussian in their nature, and occur at a high enough rate so that accidental coincidence between the two LIGO detectors is non-negligible. Glitches come in a wide range of time-frequency-amplitude morphologies, with new morphologies appearing as the detector evolves. Since they can obscure or mimic true gravitational-wave signals, a robust characterization of glitches is paramount in the effort to achieve the gravitational-wave detection rates that are predicted by the design sensitivity of LIGO. This proves a daunting task for members of the LIGO Scientific Collaboration alone due to the sheer amount of data. In this paper we describe an innovative project that combines crowdsourcing with machine learning to aid in the challenging task of categorizing all of the glitches recorded by the LIGO detectors. Through the Zooniverse platform, we engage and recruit volunteers from the public to categorize images of time-frequency representations of glitches into pre-identified morphological classes and to discover new classes that appear as the detectors evolve. In addition, machine learning algorithms are used to categorize images after being trained on human-classified examples of the morphological classes. Leveraging the strengths of both classification methods, we create a combined method with the aim of improving the efficiency and accuracy of each individual classifier. The resulting classification and characterization should help LIGO scientists to identify causes of glitches and subsequently eliminate them from the data or the detector entirely, thereby improving the rate and accuracy of gravitational-wave observations. We demonstrate these methods using a small subset of data from LIGO's first observing run.

@article{Evaggelia2017,
abstract = {Performance guarantees for recovery algorithms employed in sparse representations, and compressed sensing highlights the importance of incoherence. Optimal bounds of incoherence are attained by equiangular unit norm tight frames (ETFs). Although ETFs are important in many applications, they do not exist for all dimensions, while their construction has been proven extremely difficult. In this paper, we construct frames that are close to ETFs. According to results from frame and graph theory, the existence of an ETF depends on the existence of its signature matrix, that is, a symmetric matrix with certain structure and spectrum consisting of two distinct eigenvalues. We view the construction of a signature matrix as an inverse eigenvalue problem and propose a method that produces frames of any dimensions that are close to ETFs. Due to the achieved equiangularity property, the so obtained frames can be employed as spreading sequences in synchronous code-division multiple access (s-CDMA) systems, besides compressed sensing.},
author = {Tsiligianni, Evaggelia and Kondi, Lisimachos P. and Katsaggelos, Aggelos K.},
doi = {10.1186/s13634-017-0501-0},
issn = {1687-6180},
journal = {EURASIP Journal on Advances in Signal Processing},
keywords = {Compressed sensing,Equiangular unit norm tight frames,Signature matrix,Spreading sequences},
month = {dec},
number = {1},
pages = {66},
title = {{Approximate equiangular tight frames for compressed sensing and CDMA applications}},
url = {https://asp-eurasipjournals.springeropen.com/articles/10.1186/s13634-017-0501-0},
volume = {2017},
year = {2017}
}

Performance guarantees for recovery algorithms employed in sparse representations, and compressed sensing highlights the importance of incoherence. Optimal bounds of incoherence are attained by equiangular unit norm tight frames (ETFs). Although ETFs are important in many applications, they do not exist for all dimensions, while their construction has been proven extremely difficult. In this paper, we construct frames that are close to ETFs. According to results from frame and graph theory, the existence of an ETF depends on the existence of its signature matrix, that is, a symmetric matrix with certain structure and spectrum consisting of two distinct eigenvalues. We view the construction of a signature matrix as an inverse eigenvalue problem and propose a method that produces frames of any dimensions that are close to ETFs. Due to the achieved equiangularity property, the so obtained frames can be employed as spreading sequences in synchronous code-division multiple access (s-CDMA) systems, besides compressed sensing.

@article{Zihao2017a,
abstract = {Compressed sensing has been discussed separately in spatial and temporal domains. Compressive holography has been introduced as a method that allows 3D tomographic reconstruction at different depths from a single 2D image. Coded exposure is a temporal compressed sensing method for high speed video acquisition. In this work, we combine compressive holography and coded exposure techniques and extend the discussion to 4D reconstruction in space and time from one coded captured image. In our prototype, digital in-line holography was used for imaging macroscopic, fast moving objects. The pixel-wise temporal modulation was implemented by a digital micromirror device. In this paper we demonstrate $10\times$ temporal super resolution with multiple depths recovery from a single image. Two examples are presented for the purpose of recording subtle vibrations and tracking small particles within 5 ms.},
archivePrefix = {arXiv},
arxivId = {1610.09013},
author = {Wang, Zihao and Spinoulas, Leonidas and He, Kuan and Tian, Lei and Cossairt, Oliver and Katsaggelos, Aggelos K. and Chen, Huaijin},
doi = {10.1364/OE.25.000250},
eprint = {1610.09013},
issn = {1094-4087},
journal = {Optics Express},
month = {jan},
number = {1},
pages = {250},
pmid = {28085818},
title = {{Compressive holographic video}},
url = {https://opg.optica.org/abstract.cfm?URI=oe-25-1-250},
volume = {25},
year = {2017}
}

Compressed sensing has been discussed separately in spatial and temporal domains. Compressive holography has been introduced as a method that allows 3D tomographic reconstruction at different depths from a single 2D image. Coded exposure is a temporal compressed sensing method for high speed video acquisition. In this work, we combine compressive holography and coded exposure techniques and extend the discussion to 4D reconstruction in space and time from one coded captured image. In our prototype, digital in-line holography was used for imaging macroscopic, fast moving objects. The pixel-wise temporal modulation was implemented by a digital micromirror device. In this paper we demonstrate $10×$ temporal super resolution with multiple depths recovery from a single image. Two examples are presented for the purpose of recording subtle vibrations and tracking small particles within 5 ms.

@article{Pablo2016,
abstract = {Sequence labeling aims at assigning a label to every sample of a signal (or pixel of an image) while considering the sequentiality (or vicinity) of the samples. To perform this task, many works in the literature first filter and then label the data. Unfortunately, the filtering, which is performed independently from the labeling, is far from optimal and frequently makes the latter task harder. In this paper, a novel approach that trains a Gaussian process classifier and estimates the coefficients of an optimal filter jointly is presented. The new approach, based on Bayesian modeling and alternating direction method of multipliers (ADMMs) optimization, performs both tasks simultaneously. All unknowns are treated as stochastic variables, which are estimated using variational inference and filtering and labeling are linked with the use of ADMM. In the experimental section, synthetic and real experiments are presented to compare the proposed method with other existing approaches.},
author = {Ruiz, Pablo and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2016.2558472},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {ADMM,Bayesian Modeling,Classification,Filtering,Gaussian Processes,Variational Inference},
month = {jul},
number = {7},
pages = {3059--3072},
title = {{Joint Data Filtering and Labeling Using Gaussian Processes and Alternating Direction Method of Multipliers}},
url = {http://ieeexplore.ieee.org/document/7460233/},
volume = {25},
year = {2016}
}

Sequence labeling aims at assigning a label to every sample of a signal (or pixel of an image) while considering the sequentiality (or vicinity) of the samples. To perform this task, many works in the literature first filter and then label the data. Unfortunately, the filtering, which is performed independently from the labeling, is far from optimal and frequently makes the latter task harder. In this paper, a novel approach that trains a Gaussian process classifier and estimates the coefficients of an optimal filter jointly is presented. The new approach, based on Bayesian modeling and alternating direction method of multipliers (ADMMs) optimization, performs both tasks simultaneously. All unknowns are treated as stochastic variables, which are estimated using variational inference and filtering and labeling are linked with the use of ADMM. In the experimental section, synthetic and real experiments are presented to compare the proposed method with other existing approaches.

@article{Reza2016,
abstract = {Symmetric Nonnegative Matrix Factorization (SNMF) models arise naturally as simple reformulations of many standard clustering algorithms including the popular spectral clustering method. Recent work has demonstrated that an elementary instance of SNMF provides superior clustering quality compared to many classic clustering algorithms on a variety of synthetic and real world data sets. In this work, we present novel reformulations of this instance of SNMF based on the notion of variable splitting and produce two fast and effective algorithms for its optimization using i) the provably convergent Accelerated Proximal Gradient (APG) procedure and ii) a heuristic version of the Alternating Direction Method of Multipliers (ADMM) framework. Our two algorithms present an interesting tradeoff between computational speed and mathematical convergence guarantee: while the former method is provably convergent it is considerably slower than the latter approach, for which we also provide significant but less stringent mathematical proof regarding its convergence. Through extensive experiments we show not only that the efficacy of these approaches is equal to that of the state of the art SNMF algorithm, but also that the latter of our algorithms is extremely fast being one to two orders of magnitude faster in terms of total computation time than the state of the art approach, outperforming even spectral clustering in terms of computation time on large data sets.},
archivePrefix = {arXiv},
arxivId = {1609.05342},
author = {Borhani, Reza and Watt, Jeremy and Katsaggelos, Aggelos},
eprint = {1609.05342},
journal = {arXiv preprint arXiv:1609.05342},
month = {sep},
title = {{Fast and Effective Algorithms for Symmetric Nonnegative Matrix Factorization}},
url = {http://arxiv.org/abs/1609.05342},
year = {2016}
}

Symmetric Nonnegative Matrix Factorization (SNMF) models arise naturally as simple reformulations of many standard clustering algorithms including the popular spectral clustering method. Recent work has demonstrated that an elementary instance of SNMF provides superior clustering quality compared to many classic clustering algorithms on a variety of synthetic and real world data sets. In this work, we present novel reformulations of this instance of SNMF based on the notion of variable splitting and produce two fast and effective algorithms for its optimization using i) the provably convergent Accelerated Proximal Gradient (APG) procedure and ii) a heuristic version of the Alternating Direction Method of Multipliers (ADMM) framework. Our two algorithms present an interesting tradeoff between computational speed and mathematical convergence guarantee: while the former method is provably convergent it is considerably slower than the latter approach, for which we also provide significant but less stringent mathematical proof regarding its convergence. Through extensive experiments we show not only that the efficacy of these approaches is equal to that of the state of the art SNMF algorithm, but also that the latter of our algorithms is extremely fast being one to two orders of magnitude faster in terms of total computation time than the state of the art approach, outperforming even spectral clustering in terms of computation time on large data sets.

@article{alsaafin2016compressive,
abstract = {In this work we propose a novel framework to obtain high resolution images from compressed sensing imaging systems capturing multiple low resolution images of the same scene. The proposed approach of Compressed Sensing Super Resolution (CSSR), combines existing compressed sensing reconstruction algorithms with a low-resolution to high-resolution approach based on the use of a super Gaussian regularization term. The reconstruction alternates between compressed sensing reconstruction and super resolution reconstruction, including registration parameter estimation. The image estimation subproblem is solved using majorization-minimization while the compressed sensing reconstruction becomes an l1-minimization subject to a quadratic constraint. The performed experiments on grayscale and synthetically compressed real millimeter wave images, demonstrate the capability of the proposed framework to provide very good quality super resolved images from multiple low resolution compressed acquisitions.},
author = {AlSaafin, Wael and Villena, Salvador and Vega, Miguel and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.dsp.2015.12.005},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Compressed sensing,Image reconstruction,Passive millimeter wave images,Super resolution},
month = {mar},
pages = {180--190},
publisher = {Academic Press},
title = {{Compressive sensing super resolution from multiple observations with application to passive millimeter wave images}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200415003607},
volume = {50},
year = {2016}
}

In this work we propose a novel framework to obtain high resolution images from compressed sensing imaging systems capturing multiple low resolution images of the same scene. The proposed approach of Compressed Sensing Super Resolution (CSSR), combines existing compressed sensing reconstruction algorithms with a low-resolution to high-resolution approach based on the use of a super Gaussian regularization term. The reconstruction alternates between compressed sensing reconstruction and super resolution reconstruction, including registration parameter estimation. The image estimation subproblem is solved using majorization-minimization while the compressed sensing reconstruction becomes an l1-minimization subject to a quadratic constraint. The performed experiments on grayscale and synthetically compressed real millimeter wave images, demonstrate the capability of the proposed framework to provide very good quality super resolved images from multiple low resolution compressed acquisitions.

@article{Armin2016,
abstract = {Convolutional neural networks (CNN) are a special type of deep neural networks (DNN). They have so far been successfully applied to image super-resolution (SR) as well as other image restoration tasks. In this paper, we consider the problem of video super-resolution. We propose a CNN that is trained on both the spatial and the temporal dimensions of videos to enhance their spatial resolution. Consecutive frames are motion compensated and used as input to a CNN that provides super-resolved video frames as output. We investigate different options of combining the video frames within one CNN architecture. While large image databases are available to train deep neural networks, it is more challenging to create a large video database of sufficient quality to train neural nets for video restoration. We show that by using images to pretrain our model, a relatively small video database is sufficient for the training of our model to achieve and even improve upon the current state-of-the-art. We compare our proposed approach to current video as well as image SR algorithms.},
author = {Kappeler, Armin and Yoo, Seunghwan and Dai, Qiqin and Katsaggelos, Aggelos K.},
doi = {10.1109/TCI.2016.2532323},
issn = {2333-9403},
journal = {IEEE Transactions on Computational Imaging},
keywords = {Convolutional Neural Networks,Deep Learning,Deep Neural Networks,Video Super-Resolution},
month = {jun},
number = {2},
pages = {109--122},
title = {{Video Super-Resolution With Convolutional Neural Networks}},
url = {http://ieeexplore.ieee.org/document/7444187/},
volume = {2},
year = {2016}
}

Convolutional neural networks (CNN) are a special type of deep neural networks (DNN). They have so far been successfully applied to image super-resolution (SR) as well as other image restoration tasks. In this paper, we consider the problem of video super-resolution. We propose a CNN that is trained on both the spatial and the temporal dimensions of videos to enhance their spatial resolution. Consecutive frames are motion compensated and used as input to a CNN that provides super-resolved video frames as output. We investigate different options of combining the video frames within one CNN architecture. While large image databases are available to train deep neural networks, it is more challenging to create a large video database of sufficient quality to train neural nets for video restoration. We show that by using images to pretrain our model, a relatively small video database is sufficient for the training of our model to achieve and even improve upon the current state-of-the-art. We compare our proposed approach to current video as well as image SR algorithms.

@article{Morteza2015,
abstract = {In this paper, the performance of the phase space representation in interpreting the underlying dynamics of epileptic seizures is investigated and a novel patient-specific seizure detection approach is proposed based on the dynamics of EEG signals. To accomplish this, the trajectories of seizure and nonseizure segments are reconstructed in a high dimensional space using time-delay embedding method. Afterwards, Principal Component Analysis (PCA) was used in order to reduce the dimension of the reconstructed phase spaces. The geometry of the trajectories in the lower dimensions is then characterized using Poincar{\'{e}} section and seven features were extracted from the obtained intersection sequence. Once the features are formed, they are fed into a two-layer classification scheme, comprising the Linear Discriminant Analysis (LDA) and Naive Bayesian classifiers. The performance of the proposed method is then evaluated over the CHB-MIT benchmark database and the proposed approach achieved 88.27% sensitivity and 93.21% specificity on average with 25% training data. Finally, we perform comparative performance evaluations against the state-of-the-art methods in this domain which demonstrate the superiority of the proposed method.},
author = {Zabihi, Morteza and Kiranyaz, Serkan and Rad, Ali Bahrami and Katsaggelos, Aggelos K. and Gabbouj, Moncef and Ince, Turker},
doi = {10.1109/TNSRE.2015.2505238},
issn = {1534-4320},
journal = {IEEE Transactions on Neural Systems and Rehabilitation Engineering},
keywords = {Dynamics,Poincar{\'{e}} section,electroencephalography (EEG),phase space,seizure detection,two-layer classifier topology},
month = {mar},
number = {3},
pages = {386--398},
pmid = {26701865},
title = {{Analysis of High-Dimensional Phase Space via Poincar{\'{e}} Section for Patient-Specific Seizure Detection}},
url = {https://ieeexplore.ieee.org/document/7360936/},
volume = {24},
year = {2016}
}

In this paper, the performance of the phase space representation in interpreting the underlying dynamics of epileptic seizures is investigated and a novel patient-specific seizure detection approach is proposed based on the dynamics of EEG signals. To accomplish this, the trajectories of seizure and nonseizure segments are reconstructed in a high dimensional space using time-delay embedding method. Afterwards, Principal Component Analysis (PCA) was used in order to reduce the dimension of the reconstructed phase spaces. The geometry of the trajectories in the lower dimensions is then characterized using Poincaré section and seven features were extracted from the obtained intersection sequence. Once the features are formed, they are fed into a two-layer classification scheme, comprising the Linear Discriminant Analysis (LDA) and Naive Bayesian classifiers. The performance of the proposed method is then evaluated over the CHB-MIT benchmark database and the proposed approach achieved 88.27% sensitivity and 93.21% specificity on average with 25% training data. Finally, we perform comparative performance evaluations against the state-of-the-art methods in this domain which demonstrate the superiority of the proposed method.

@article{AliBahrami2016,
abstract = {Aim: Resuscitation guidelines recommend different treatments depending on the patient's cardiac rhythm. Rhythm interpretation is a key tool to retrospectively evaluate and improve the quality of treatment. Manual rhythm annotation is time consuming and an obstacle for handling large resuscitation datasets efficiently. The objective of this study was to develop a system for automatic rhythm interpretation by using signal processing and machine learning algorithms. Methods: Data from 302 out of hospital cardiac arrest patients were used. In total 1669 3-second artifact free ECG segments with clinical rhythm annotations were extracted. The proposed algorithms combine 32 features obtained from both wavelet- and time-domain representations of the ECG, followed by a feature selection procedure based on the wrapper method in a nested cross-validation architecture. Linear and quadratic discriminant analyses (LDA and QDA) were used to automatically classify the segments into one of five rhythm types: ventricular tachycardia (VT), ventricular fibrillation (VF), pulseless electrical activity (PEA), asystole (AS), and pulse generating rhythms (PR). Results: The overall accuracy for the best algorithm was 68%. VT, VF, and AS are recognized with sensitivities of 71%, 75%, and 79%, respectively. Sensitivities for PEA and PR were 55% and 56%, respectively, which reflects the difficulty of identifying pulse using only the ECG. Conclusions: An ECG based automatic rhythm interpreter for resuscitation has been demonstrated. The interpreter handles VT, VF and AS well, while PEA and PR discrimination poses a more difficult problem.},
author = {Rad, Ali Bahrami and Engan, Kjersti and Katsaggelos, Aggelos K. and Kval{\o}y, Jan Terje and Wik, Lars and Kramer-Johansen, Jo and Irusta, Unai and Eftest{\o}l, Trygve},
doi = {10.1016/j.resuscitation.2016.01.015},
issn = {03009572},
journal = {Resuscitation},
keywords = {Cardiac rhythm interpretation,Cardiopulmonary resuscitation,Classification,Feature extraction,Feature selection},
month = {may},
pages = {44--50},
pmid = {26891862},
title = {{Automatic cardiac rhythm interpretation during resuscitation}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0300957216000411},
volume = {102},
year = {2016}
}

Aim: Resuscitation guidelines recommend different treatments depending on the patient's cardiac rhythm. Rhythm interpretation is a key tool to retrospectively evaluate and improve the quality of treatment. Manual rhythm annotation is time consuming and an obstacle for handling large resuscitation datasets efficiently. The objective of this study was to develop a system for automatic rhythm interpretation by using signal processing and machine learning algorithms. Methods: Data from 302 out of hospital cardiac arrest patients were used. In total 1669 3-second artifact free ECG segments with clinical rhythm annotations were extracted. The proposed algorithms combine 32 features obtained from both wavelet- and time-domain representations of the ECG, followed by a feature selection procedure based on the wrapper method in a nested cross-validation architecture. Linear and quadratic discriminant analyses (LDA and QDA) were used to automatically classify the segments into one of five rhythm types: ventricular tachycardia (VT), ventricular fibrillation (VF), pulseless electrical activity (PEA), asystole (AS), and pulse generating rhythms (PR). Results: The overall accuracy for the best algorithm was 68%. VT, VF, and AS are recognized with sensitivities of 71%, 75%, and 79%, respectively. Sensitivities for PEA and PR were 55% and 56%, respectively, which reflects the difficulty of identifying pulse using only the ECG. Conclusions: An ECG based automatic rhythm interpreter for resuscitation has been demonstrated. The interpreter handles VT, VF and AS well, while PEA and PR discrimination poses a more difficult problem.

@article{Xiang2016,
abstract = {Because art is inherently visual, the use of imaging has long been an important way to understand its structure, form, and history. Recently, new ways of engaging with objects from our shared cultural heritage are possible with advances in computation and imaging that allow scientists to analyze art noninvasively, historians to pose new social questions about the art, and the public to explore and interact with art in ways never before possible. There is a rich history in applying image processing techniques to conventional photographic images of works of art, many of which have been highlighted in previous special issues of IEEE Signal Processing Magazine (e.g., the 2008 and 2015 July issues). Building on these contributions, this article comprises a survey of techniques where computation is central to the image acquisition process. Known as computational imaging, the methods being pioneered in this field are increasingly relevant to cultural heritage applications because they leverage advances in image processing, acquisition, and display technologies that make scientific data readily comprehensible to a broad cohort of nontechnical researchers interested in understanding the visual content of art. Presently, only a small research community undertakes computational imaging of cultural heritage. Here we aim to introduce this growing new field to a larger research community by discussing: 1) the historic background of imaging of art, 2) the burgeoning present day community of researchers interested in computational imaging in the arts, and finally, 3) our vision for the future of this new field.},
author = {Huang, Xiang and Uffelman, Erich and Cossairt, Oliver and Walton, Marc and Katsaggelos, Aggelos K.},
doi = {10.1109/MSP.2016.2581847},
issn = {1053-5888},
journal = {IEEE Signal Processing Magazine},
month = {sep},
number = {5},
pages = {130--138},
title = {{Computational Imaging for Cultural Heritage: Recent developments in spectral imaging, 3-D surface measurement, image relighting, and X-ray mapping}},
url = {http://ieeexplore.ieee.org/document/7560020/},
volume = {33},
year = {2016}
}

Because art is inherently visual, the use of imaging has long been an important way to understand its structure, form, and history. Recently, new ways of engaging with objects from our shared cultural heritage are possible with advances in computation and imaging that allow scientists to analyze art noninvasively, historians to pose new social questions about the art, and the public to explore and interact with art in ways never before possible. There is a rich history in applying image processing techniques to conventional photographic images of works of art, many of which have been highlighted in previous special issues of IEEE Signal Processing Magazine (e.g., the 2008 and 2015 July issues). Building on these contributions, this article comprises a survey of techniques where computation is central to the image acquisition process. Known as computational imaging, the methods being pioneered in this field are increasingly relevant to cultural heritage applications because they leverage advances in image processing, acquisition, and display technologies that make scientific data readily comprehensible to a broad cohort of nontechnical researchers interested in understanding the visual content of art. Presently, only a small research community undertakes computational imaging of cultural heritage. Here we aim to introduce this growing new field to a larger research community by discussing: 1) the historic background of imaging of art, 2) the burgeoning present day community of researchers interested in computational imaging in the arts, and finally, 3) our vision for the future of this new field.

@article{Roman2015,
abstract = {We present a prototype compressive video camera that encodes scene movement using a translated binary photomask in the optical path. The encoded recording can then be used to reconstruct multiple output frames from each captured image, effectively synthesizing high speed video. The use of a printed binary mask allows reconstruction at higher spatial resolutions than has been previously demonstrated. In addition, we improve upon previous work by investigating tradeoffs in mask design and reconstruction algorithm selection. We identify a mask design that consistently provides the best performance across multiple reconstruction strategies in simulation, and verify it with our prototype hardware. Finally, we compare reconstruction algorithms and identify the best choice in terms of balancing reconstruction quality and speed.},
author = {Koller, Roman and Schmid, Lukas and Matsuda, Nathan and Niederberger, Thomas and Spinoulas, Leonidas and Cossairt, Oliver and Schuster, Guido and Katsaggelos, Aggelos K.},
doi = {10.1364/OE.23.015992},
issn = {1094-4087},
journal = {Optics Express},
month = {jun},
number = {12},
pages = {15992},
title = {{High spatio-temporal resolution video with compressed sensing}},
url = {https://opg.optica.org/abstract.cfm?URI=oe-23-12-15992},
volume = {23},
year = {2015}
}

We present a prototype compressive video camera that encodes scene movement using a translated binary photomask in the optical path. The encoded recording can then be used to reconstruct multiple output frames from each captured image, effectively synthesizing high speed video. The use of a printed binary mask allows reconstruction at higher spatial resolutions than has been previously demonstrated. In addition, we improve upon previous work by investigating tradeoffs in mask design and reconstruction algorithm selection. We identify a mask design that consistently provides the best performance across multiple reconstruction strategies in simulation, and verify it with our prototype hardware. Finally, we compare reconstruction algorithms and identify the best choice in terms of balancing reconstruction quality and speed.

@article{Evaggelia2015,
abstract = {Performance guarantees for the algorithms deployed to solve underdetermined linear systems with sparse solutions are based on the assumption that the involved system matrix has the form of an incoherent unit norm tight frame. Learned dictionaries, which are popular in sparse representations, often do not meet the necessary conditions for signal recovery. In compressed sensing (CS), recovery rates have been improved substantially with optimized projections; however, these techniques do not produce binary matrices, which are more suitable for hardware implementation. In this paper, we consider an underdetermined linear system with sparse solutions and propose a preconditioning technique that yields a system matrix having the properties of an incoherent unit norm tight frame. While existing work in preconditioning concerns greedy algorithms, the proposed technique is based on recent theoretical results for standard numerical solvers such as BP and OMP. Our simulations show that the proposed preconditioning improves the recovery rates both in sparse representations and CS; the results for CS are comparable to optimized projections.},
author = {Tsiligianni, Evaggelia and Kondi, Lisimachos P. and Katsaggelos, Aggelos K.},
doi = {10.1109/LSP.2015.2392000},
issn = {1070-9908},
journal = {IEEE Signal Processing Letters},
keywords = {Compressed sensing,incoherent unit norm tight frames,preconditioning,sparse representations},
month = {sep},
number = {9},
pages = {1239--1243},
title = {{Preconditioning for Underdetermined Linear Systems with Sparse Solutions}},
url = {http://ieeexplore.ieee.org/document/7008458/},
volume = {22},
year = {2015}
}

Performance guarantees for the algorithms deployed to solve underdetermined linear systems with sparse solutions are based on the assumption that the involved system matrix has the form of an incoherent unit norm tight frame. Learned dictionaries, which are popular in sparse representations, often do not meet the necessary conditions for signal recovery. In compressed sensing (CS), recovery rates have been improved substantially with optimized projections; however, these techniques do not produce binary matrices, which are more suitable for hardware implementation. In this paper, we consider an underdetermined linear system with sparse solutions and propose a preconditioning technique that yields a system matrix having the properties of an incoherent unit norm tight frame. While existing work in preconditioning concerns greedy algorithms, the proposed technique is based on recent theoretical results for standard numerical solvers such as BP and OMP. Our simulations show that the proposed preconditioning improves the recovery rates both in sparse representations and CS; the results for CS are comparable to optimized projections.

@article{chen2015robust,
abstract = {In this paper, we consider the problem of recovering jointly sparse vectors from underdetermined measurements that are corrupted by both additive noise and outliers. This can be viewed as the robust extension of the Multiple Measurement Vector (MMV) problem. To solve this problem, we propose two general approaches. As a benchmark, the first approach preprocesses the input for outlier removal and then employs state-of-the-art technologies for signal recovery. The second approach, as the main contribution of this paper, is based on formulation of an innovative regularized fitting problem. By solving the regularized fitting problem, we jointly remove outliers and recover the sparse vectors. Furthermore, by exploiting temporal smoothness among the sparse vectors, we improve noise robustness of the proposed approach and avoid the problem of over-fitting. Extensive numerical results are provided to illustrate the excellent performance of the proposed approach.},
author = {Chen, Zhaofu and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TSP.2015.2403277},
issn = {1053-587X},
journal = {IEEE Transactions on Signal Processing},
keywords = {Signal reconstruction,iterative methods,optimization},
month = {apr},
number = {7},
pages = {1779--1791},
publisher = {IEEE},
title = {{Robust Recovery of Temporally Smooth Signals From Under-Determined Multiple Measurements}},
url = {http://ieeexplore.ieee.org/document/7041206/},
volume = {63},
year = {2015}
}

In this paper, we consider the problem of recovering jointly sparse vectors from underdetermined measurements that are corrupted by both additive noise and outliers. This can be viewed as the robust extension of the Multiple Measurement Vector (MMV) problem. To solve this problem, we propose two general approaches. As a benchmark, the first approach preprocesses the input for outlier removal and then employs state-of-the-art technologies for signal recovery. The second approach, as the main contribution of this paper, is based on formulation of an innovative regularized fitting problem. By solving the regularized fitting problem, we jointly remove outliers and recover the sparse vectors. Furthermore, by exploiting temporal smoothness among the sparse vectors, we improve noise robustness of the proposed approach and avoid the problem of over-fitting. Extensive numerical results are provided to illustrate the excellent performance of the proposed approach.

@article{Xu2015,
abstract = {Blind image deconvolution involves two key objectives: 1) latent image and 2) blur estimation. For latent image estimation, we propose a fast deconvolution algorithm, which uses an image prior of nondimensional Gaussianity measure to enforce sparsity and an undetermined boundary condition methodology to reduce boundary artifacts. For blur estimation, a linear inverse problem with normalization and nonnegative constraints must be solved. However, the normalization constraint is ignored in many blind image deblurring methods, mainly because it makes the problem less tractable. In this paper, we show that the normalization constraint can be very naturally incorporated into the estimation process by using a Dirichlet distribution to approximate the posterior distribution of the blur. Making use of variational Dirichlet approximation, we provide a blur posterior approximation that considers the uncertainty of the estimate and removes noise in the estimated kernel. Experiments with synthetic and real data demonstrate that the proposed method is very competitive to the state-of-the-art blind image restoration methods.},
author = {Zhou, Xu and Mateos, Javier and Zhou, Fugen and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2015.2478407},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Blind deconvolution,Dirichlet distribution,constrained optimization,image deblurring,inverse problem,point spread function,variational distribution approximations},
month = {dec},
number = {12},
pages = {5127--5139},
title = {{Variational Dirichlet Blur Kernel Estimation}},
url = {http://ieeexplore.ieee.org/document/7265038/},
volume = {24},
year = {2015}
}

Blind image deconvolution involves two key objectives: 1) latent image and 2) blur estimation. For latent image estimation, we propose a fast deconvolution algorithm, which uses an image prior of nondimensional Gaussianity measure to enforce sparsity and an undetermined boundary condition methodology to reduce boundary artifacts. For blur estimation, a linear inverse problem with normalization and nonnegative constraints must be solved. However, the normalization constraint is ignored in many blind image deblurring methods, mainly because it makes the problem less tractable. In this paper, we show that the normalization constraint can be very naturally incorporated into the estimation process by using a Dirichlet distribution to approximate the posterior distribution of the blur. Making use of variational Dirichlet approximation, we provide a blur posterior approximation that considers the uncertainty of the estimate and removes noise in the estimated kernel. Experiments with synthetic and real data demonstrate that the proposed method is very competitive to the state-of-the-art blind image restoration methods.

@article{Aggelos2015,
abstract = {In this paper, we review recent results on audiovisual (AV) fusion. We also discuss some of the challenges and report on approaches to address them. One important issue in AV fusion is how the modalities interact and influence each other. This review will address this question in the context of AV speech processing, and especially speech recognition, where one of the issues is that the modalities both interact but also sometimes appear to desynchronize from each other. An additional issue that sometimes arises is that one of the modalities may be missing at test time, although it is available at training time; for example, it may be possible to collect AV training data while only having access to audio at test time. We will review approaches to address this issue from the area of multiview learning, where the goal is to learn a model or representation for each of the modalities separately while taking advantage of the rich multimodal training data. In addition to multiview learning, we also discuss the recent application of deep learning (DL) toward AV fusion. We finally draw conclusions and offer our assessment of the future in the area of AV fusion.},
author = {Katsaggelos, Aggelos K. and Bahaadini, Sara and Molina, Rafael},
doi = {10.1109/JPROC.2015.2459017},
issn = {0018-9219},
journal = {Proceedings of the IEEE},
keywords = {Audiovisual (AV) fusion,deep learning (DL),machine learning,multimodal analysis,multiview learning,stream asynchrony},
month = {sep},
number = {9},
pages = {1635--1653},
title = {{Audiovisual Fusion: Challenges and New Approaches}},
url = {http://ieeexplore.ieee.org/document/7194741/},
volume = {103},
year = {2015}
}

In this paper, we review recent results on audiovisual (AV) fusion. We also discuss some of the challenges and report on approaches to address them. One important issue in AV fusion is how the modalities interact and influence each other. This review will address this question in the context of AV speech processing, and especially speech recognition, where one of the issues is that the modalities both interact but also sometimes appear to desynchronize from each other. An additional issue that sometimes arises is that one of the modalities may be missing at test time, although it is available at training time; for example, it may be possible to collect AV training data while only having access to audio at test time. We will review approaches to address this issue from the area of multiview learning, where the goal is to learn a model or representation for each of the modalities separately while taking advantage of the rich multimodal training data. In addition to multiview learning, we also discuss the recent application of deep learning (DL) toward AV fusion. We finally draw conclusions and offer our assessment of the future in the area of AV fusion.

@article{Nicolas2015,
abstract = {Yan is an ETS-domain transcription factor responsible for maintaining Drosophila eye cells in a multipotent state. Yan is at the core of a regulatory network that determines the time and place in which cells transit from multipotency to one of several differentiated lineages. Using a fluorescent reporter for Yan expression, we observed a biphasic distribution of Yan in multipotent cells, with a rapid inductive phase and slow decay phase. Transitions to various differentiated states occurred over the course of this dynamic process, suggesting that Yan expression level does not strongly determine cell potential. Consistent with this conclusion, perturbing Yan expression by varying gene dosage had no effect on cell fate transitions. However, we observed that as cells transited to differentiation, Yan expression became highly heterogeneous and this heterogeneity was transient. Signals received via the EGF Receptor were necessary for the transience in Yan noise since genetic loss caused sustained noise. Since these signals are essential for eye cells to differentiate, we suggest that dynamic heterogeneity of Yan is a necessary element of the transition process, and cell states are stabilized through noise reduction.},
author = {Pel{\'{a}}ez, Nicol{\'{a}}s and Gavalda-Miralles, Arnau and Wang, Bao and Navarro, Heliodoro Tejedor and Gudjonson, Herman and Rebay, Ilaria and Dinner, Aaron R. and Katsaggelos, Aggelos K. and Amaral, Lu{\'{i}}s A.N. and Carthew, Richard W.},
doi = {10.7554/eLife.08924},
issn = {2050-084X},
journal = {eLife},
month = {nov},
number = {NOVEMBER2015},
pmid = {26583752},
title = {{Dynamics and heterogeneity of a fate determinant during transition towards cell differentiation}},
url = {https://elifesciences.org/articles/08924},
volume = {4},
year = {2015}
}

Yan is an ETS-domain transcription factor responsible for maintaining Drosophila eye cells in a multipotent state. Yan is at the core of a regulatory network that determines the time and place in which cells transit from multipotency to one of several differentiated lineages. Using a fluorescent reporter for Yan expression, we observed a biphasic distribution of Yan in multipotent cells, with a rapid inductive phase and slow decay phase. Transitions to various differentiated states occurred over the course of this dynamic process, suggesting that Yan expression level does not strongly determine cell potential. Consistent with this conclusion, perturbing Yan expression by varying gene dosage had no effect on cell fate transitions. However, we observed that as cells transited to differentiation, Yan expression became highly heterogeneous and this heterogeneity was transient. Signals received via the EGF Receptor were necessary for the transience in Yan noise since genetic loss caused sustained noise. Since these signals are essential for eye cells to differentiate, we suggest that dynamic heterogeneity of Yan is a necessary element of the transition process, and cell states are stabilized through noise reduction.

@article{LasyaPriya2015,
abstract = {Introduction: Patients surviving myocardial infarction (MI) can be divided into high and low arrhythmic risk groups. Distinguishing between these two groups is of crucial importance since the high-risk group has been shown to benefit from implantable cardioverter defibrillator insertion; a costly surgical procedure with potential complications and no proven advantages for the low-risk group. Currently, markers such as left ventricular ejection fraction and myocardial scar size are used to evaluate arrhythmic risk. Methods: In this paper, we propose quantitative discriminative features extracted from late gadolinium enhanced cardiac magnetic resonance images of post-MI patients, to distinguish between 20 high-risk and 34 low-risk patients. These features include size, location, and textural information concerning the scarred myocardium. To evaluate the discriminative power of the proposed features, we used several built-in classification schemes from matrix laboratory (MATLAB) and Waikato environment for knowledge analysis (WEKA) software, including k-nearest neighbor (k-NN), support vector machine (SVM), decision tree, and random forest. Results: In Experiment 1, the leave-one-out cross-validation scheme is implemented in MATLAB to classify high- and low-risk groups with a classification accuracy of 94.44%, and an AUC of 0.965 for a feature combination that captures size, location and heterogeneity of the scar. In Experiment 2 with the help of WEKA, nested cross-validation is performed with k-NN, SVM, adjusting decision tree and random forest classifiers to differentiate high-risk and low-risk patients. SVM classifier provided average accuracy of 92.6%, and AUC of 0.921 for a feature combination capturing location and heterogeneity of the scar. Experiment 1 and Experiment 2 show that textural features from the scar are important for classification and that localization features provide an additional benefit. Conclusion: These promising results suggest that the discriminative features introduced in this paper can be used by medical professionals, or in automatic decision support systems, along with the recognized risk markers, to improve arrhythmic risk stratification in post-MI patients.},
author = {Kotu, Lasya Priya and Engan, Kjersti and Borhani, Reza and Katsaggelos, Aggelos K. and {\O}rn, Stein and Woie, Leik and Eftest{\o}l, Trygve},
doi = {10.1016/j.artmed.2015.06.001},
issn = {09333657},
journal = {Artificial Intelligence in Medicine},
keywords = {Cardiac magnetic resonance image,High and low arrhythmic risk,K-Nearest neighbor classifier,Local binary pattern,Sobel filter,Support vector machine classifier},
month = {jul},
number = {3},
pages = {205--215},
pmid = {26239472},
title = {{Cardiac magnetic resonance image-based classification of the risk of arrhythmias in post-myocardial infarction patients}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0933365715000809},
volume = {64},
year = {2015}
}

Introduction: Patients surviving myocardial infarction (MI) can be divided into high and low arrhythmic risk groups. Distinguishing between these two groups is of crucial importance since the high-risk group has been shown to benefit from implantable cardioverter defibrillator insertion; a costly surgical procedure with potential complications and no proven advantages for the low-risk group. Currently, markers such as left ventricular ejection fraction and myocardial scar size are used to evaluate arrhythmic risk. Methods: In this paper, we propose quantitative discriminative features extracted from late gadolinium enhanced cardiac magnetic resonance images of post-MI patients, to distinguish between 20 high-risk and 34 low-risk patients. These features include size, location, and textural information concerning the scarred myocardium. To evaluate the discriminative power of the proposed features, we used several built-in classification schemes from matrix laboratory (MATLAB) and Waikato environment for knowledge analysis (WEKA) software, including k-nearest neighbor (k-NN), support vector machine (SVM), decision tree, and random forest. Results: In Experiment 1, the leave-one-out cross-validation scheme is implemented in MATLAB to classify high- and low-risk groups with a classification accuracy of 94.44%, and an AUC of 0.965 for a feature combination that captures size, location and heterogeneity of the scar. In Experiment 2 with the help of WEKA, nested cross-validation is performed with k-NN, SVM, adjusting decision tree and random forest classifiers to differentiate high-risk and low-risk patients. SVM classifier provided average accuracy of 92.6%, and AUC of 0.921 for a feature combination capturing location and heterogeneity of the scar. Experiment 1 and Experiment 2 show that textural features from the scar are important for classification and that localization features provide an additional benefit. Conclusion: These promising results suggest that the discriminative features introduced in this paper can be used by medical professionals, or in automatic decision support systems, along with the recognized risk markers, to improve arrhythmic risk stratification in post-MI patients.

@article{Matthew2015,
abstract = {The image processing technique known as superresolution (SR) has the potential to allow engineers to specify lower resolution and, therefore, less expensive cameras for a given task by enhancing the base camera's resolution. This is especially true in the remote detection and classification of objects in the environment, such as aircraft or human faces. Performing each of these tasks requires a minimum image “sharpness” which is quantified by a maximum resolvable spatial frequency, which is, in turn, a function of the camera optics, pixel sampling density, and signal-to-noise ratio. Much of the existing SR literature focuses on SR performance metrics for candidate algorithms, such as perceived image quality or peak SNR. These metrics can be misleading because they also credit deblurring and/or denoising in addition to true SR. In this paper, we propose a new, task-based metric where the performance of an SR algorithm is, instead, directly tied to the probability of successfully detecting critical spatial frequencies within the scene.},
author = {Woods, Matthew and Katsaggelos, Aggelos K.},
doi = {10.1364/JOSAA.32.002002},
issn = {1084-7529},
journal = {Journal of the Optical Society of America A},
month = {nov},
number = {11},
pages = {2002},
title = {{Spatial-frequency-based metric for image superresolution}},
url = {https://opg.optica.org/abstract.cfm?URI=josaa-32-11-2002},
volume = {32},
year = {2015}
}

The image processing technique known as superresolution (SR) has the potential to allow engineers to specify lower resolution and, therefore, less expensive cameras for a given task by enhancing the base camera's resolution. This is especially true in the remote detection and classification of objects in the environment, such as aircraft or human faces. Performing each of these tasks requires a minimum image “sharpness” which is quantified by a maximum resolvable spatial frequency, which is, in turn, a function of the camera optics, pixel sampling density, and signal-to-noise ratio. Much of the existing SR literature focuses on SR performance metrics for candidate algorithms, such as perceived image quality or peak SNR. These metrics can be misleading because they also credit deblurring and/or denoising in addition to true SR. In this paper, we propose a new, task-based metric where the performance of an SR algorithm is, instead, directly tied to the probability of successfully detecting critical spatial frequencies within the scene.

@article{Bruno2013,
abstract = {In this paper we provide a review of the recent literature on Bayesian Blind Image Deconvolution (BID) methods. We believe that two events have marked the recent history of BID: the predominance of Variational Bayes (VB) inference as a tool to solve BID problems and the increasing interest of the computer vision community in solving BID problems. VB inference in combination with recent image models like the ones based on Super Gaussian (SG) and Scale Mixture of Gaussians (SMG) representations have led to the use of very general and powerful tools to provide clear images from blurry observations. In the provided review emphasis is paid on VB inference and the use of SG and SMG models with coverage of recent advances in sampling methods. We also provide examples of current state of the art BID methods and discuss problems that very likely will mark the near future of BID.},
author = {Ruiz, Pablo and Zhou, Xu and Mateos, Javier and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.dsp.2015.04.012},
isbn = {9780992862602 22195491 , issue = 10},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Bayesian modeling,Blind deconvolution,Image deblurring,Image restoration,Variational Bayesian},
month = {dec},
pages = {116--127},
title = {{Variational Bayesian Blind Image Deconvolution: A review}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S105120041500144X},
volume = {47},
year = {2015}
}

In this paper we provide a review of the recent literature on Bayesian Blind Image Deconvolution (BID) methods. We believe that two events have marked the recent history of BID: the predominance of Variational Bayes (VB) inference as a tool to solve BID problems and the increasing interest of the computer vision community in solving BID problems. VB inference in combination with recent image models like the ones based on Super Gaussian (SG) and Scale Mixture of Gaussians (SMG) representations have led to the use of very general and powerful tools to provide clear images from blurry observations. In the provided review emphasis is paid on VB inference and the use of SG and SMG models with coverage of recent advances in sampling methods. We also provide examples of current state of the art BID methods and discuss problems that very likely will mark the near future of BID.

@article{Pablo2013a,
abstract = {In recent years, kernel methods, in particular support vector machines (SVMs), have been successfully introduced to remote sensing image classification. Their properties make them appropriate for dealing with a high number of image features and a low number of available labeled spectra. The introduction of alternative approaches based on (parametric) Bayesian inference has been quite scarce in the more recent years. Assuming a particular prior data distribution may lead to poor results in remote sensing problems because of the specificities and complexity of the data. In this context, the emerging field of nonparametric Bayesian methods constitutes a proper theoretical framework to tackle the remote sensing image classification problem. This paper exploits the Bayesian modeling and inference paradigm to tackle the problem of kernel-based remote sensing image classification. This Bayesian methodology is appropriate for both finite- and infinite-dimensional feature spaces. The particular problem of active learning is addressed by proposing an incremental/active learning approach based on three different approaches: 1) the maximum differential of entropies; 2) the minimum distance to decision boundary; and 3) the minimum normalized distance. Parameters are estimated by using the evidence Bayesian approach, the kernel trick, and the marginal distribution of the observations instead of the posterior distribution of the adaptive parameters. This approach allows us to deal with infinite-dimensional feature spaces. The proposed approach is tested on the challenging problem of urban monitoring from multispectral and synthetic aperture radar data and in multiclass land cover classification of hyperspectral images, in both purely supervised and active learning settings. Similar results are obtained when compared to SVMs in the supervised mode, with the advantage of providing posterior estimates for classification and automatic parameter learning. Comparison with random sampling as well as standard active learning methods such as margin sampling and entropy-query-by-bagging reveals a systematic overall accuracy gain and faster convergence with the number of queries. {\textcopyright} 2013 IEEE.},
author = {Ruiz, Pablo and Mateos, Javier and Camps-Valls, Gustavo and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TGRS.2013.2258468},
issn = {0196-2892},
journal = {IEEE Transactions on Geoscience and Remote Sensing},
keywords = {Bayesian inference,Incremental/active learning,Multispectral image segmentation,Supervised classification},
month = {apr},
number = {4},
pages = {2186--2196},
title = {{Bayesian Active Remote Sensing Image Classification}},
url = {http://ieeexplore.ieee.org/document/6521357/},
volume = {52},
year = {2014}
}

In recent years, kernel methods, in particular support vector machines (SVMs), have been successfully introduced to remote sensing image classification. Their properties make them appropriate for dealing with a high number of image features and a low number of available labeled spectra. The introduction of alternative approaches based on (parametric) Bayesian inference has been quite scarce in the more recent years. Assuming a particular prior data distribution may lead to poor results in remote sensing problems because of the specificities and complexity of the data. In this context, the emerging field of nonparametric Bayesian methods constitutes a proper theoretical framework to tackle the remote sensing image classification problem. This paper exploits the Bayesian modeling and inference paradigm to tackle the problem of kernel-based remote sensing image classification. This Bayesian methodology is appropriate for both finite- and infinite-dimensional feature spaces. The particular problem of active learning is addressed by proposing an incremental/active learning approach based on three different approaches: 1) the maximum differential of entropies; 2) the minimum distance to decision boundary; and 3) the minimum normalized distance. Parameters are estimated by using the evidence Bayesian approach, the kernel trick, and the marginal distribution of the observations instead of the posterior distribution of the adaptive parameters. This approach allows us to deal with infinite-dimensional feature spaces. The proposed approach is tested on the challenging problem of urban monitoring from multispectral and synthetic aperture radar data and in multiclass land cover classification of hyperspectral images, in both purely supervised and active learning settings. Similar results are obtained when compared to SVMs in the supervised mode, with the advantage of providing posterior estimates for classification and automatic parameter learning. Comparison with random sampling as well as standard active learning methods such as margin sampling and entropy-query-by-bagging reveals a systematic overall accuracy gain and faster convergence with the number of queries. © 2013 IEEE.

@article{Lei2014,
abstract = {Our work addresses the problem of analyzing and understanding dynamic video scenes. A two-level motion pattern mining approach is proposed. At the first level, activities are modeled as distributions over patch-based features, including spatial location, moving direction, and speed. At the second level, traffic states are modeled as distributions over activities. Both patterns are shared among video clips. Compared to other works, one advantage of our method is that moving speed is considered to describe visual word. The other advantage is that traffic states are detected and assigned to every video frame. These enable finer semantic interpretation, more precise video segmentation, and anomaly detection. Specifically, every video frame is labeled by a certain traffic state, and the video is segmented frame by frame accordingly. Moving pixels in each frame, which do not belong to any activity or cannot exist in the corresponding traffic state, are detected as anomalies. We have successfully tested our approach on some challenging traffic surveillance sequences containing both pedestrian and vehicle motions. {\textcopyright} 2000-2011 IEEE.},
author = {Song, Lei and Jiang, Fan and Shi, Zhongke and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TITS.2014.2299403},
issn = {1524-9050},
journal = {IEEE Transactions on Intelligent Transportation Systems},
keywords = {Anomaly detection,Latent Dirichlet Allocation (LDA),motion pattern analysis,video segmentation,visual surveillance},
month = {jun},
number = {3},
pages = {1273--1285},
title = {{Toward Dynamic Scene Understanding by Hierarchical Motion Pattern Mining}},
url = {https://ieeexplore.ieee.org/document/6746216/},
volume = {15},
year = {2014}
}

Our work addresses the problem of analyzing and understanding dynamic video scenes. A two-level motion pattern mining approach is proposed. At the first level, activities are modeled as distributions over patch-based features, including spatial location, moving direction, and speed. At the second level, traffic states are modeled as distributions over activities. Both patterns are shared among video clips. Compared to other works, one advantage of our method is that moving speed is considered to describe visual word. The other advantage is that traffic states are detected and assigned to every video frame. These enable finer semantic interpretation, more precise video segmentation, and anomaly detection. Specifically, every video frame is labeled by a certain traffic state, and the video is segmented frame by frame accordingly. Moving pixels in each frame, which do not belong to any activity or cannot exist in the corresponding traffic state, are detected as anomalies. We have successfully tested our approach on some challenging traffic surveillance sequences containing both pedestrian and vehicle motions. © 2000-2011 IEEE.

@article{chen2014variational,
abstract = {In this paper we present an introduction to Variational Bayesian (VB) methods in the context of probabilistic graphical models, and discuss their application in multimedia related problems. VB is a family of deterministic probability distribution approximation procedures that offer distinct advantages over alternative approaches based on stochastic sampling and those providing only point estimates. VB inference is flexible to be applied in different practical problems, yet is broad enough to subsume as its special cases several alternative inference approaches including Maximum A Posteriori (MAP) and the Expectation-Maximization (EM) algorithm. In this paper we also show the connections between VB and other posterior approximation methods such as the marginalization-based Loopy Belief Propagation (LBP) and the Expectation Propagation (EP) algorithms. Specifically, both VB and EP are variational methods that minimize functionals based on the Kullback-Leibler (KL) divergence. LBP, traditionally developed using graphical models, can also be viewed as a VB inference procedure. We present several multimedia related applications illustrating the use and effectiveness of the VB algorithms discussed herein. We hope that by reading this tutorial the readers will obtain a general understanding of Bayesian methods and establish connections among popular algorithms used in practice. {\textcopyright} 1999-2012 IEEE.},
author = {Chen, Zhaofu and Babacan, S. Derin and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TMM.2014.2307692},
issn = {1520-9210},
journal = {IEEE Transactions on Multimedia},
keywords = {Bayes methods,graphical models,inverse problems,multimedia signal processing,variational Bayes},
month = {jun},
number = {4},
pages = {1000--1017},
publisher = {IEEE},
title = {{Variational Bayesian Methods For Multimedia Problems}},
url = {http://ieeexplore.ieee.org/document/6747301/},
volume = {16},
year = {2014}
}

In this paper we present an introduction to Variational Bayesian (VB) methods in the context of probabilistic graphical models, and discuss their application in multimedia related problems. VB is a family of deterministic probability distribution approximation procedures that offer distinct advantages over alternative approaches based on stochastic sampling and those providing only point estimates. VB inference is flexible to be applied in different practical problems, yet is broad enough to subsume as its special cases several alternative inference approaches including Maximum A Posteriori (MAP) and the Expectation-Maximization (EM) algorithm. In this paper we also show the connections between VB and other posterior approximation methods such as the marginalization-based Loopy Belief Propagation (LBP) and the Expectation Propagation (EP) algorithms. Specifically, both VB and EP are variational methods that minimize functionals based on the Kullback-Leibler (KL) divergence. LBP, traditionally developed using graphical models, can also be viewed as a VB inference procedure. We present several multimedia related applications illustrating the use and effectiveness of the VB algorithms discussed herein. We hope that by reading this tutorial the readers will obtain a general understanding of Bayesian methods and establish connections among popular algorithms used in practice. © 1999-2012 IEEE.

@article{Salvador2014,
abstract = {A new Bayesian Super-Resolution (SR) image registration and reconstruction method is proposed. The new method utilizes a prior distribution based on a general combination of spatially adaptive, or non-stationary, image filters, which includes an adaptive local strength parameter able to preserve both image edges and textures. With the application of variational techniques, the proposed method allows for the automatic estimation of all problem unknowns. An experimental comparison between state of the art methods and the proposed SR approach has been performed on both synthetic and real images. {\textcopyright} 2014 Elsevier Inc.},
author = {Villena, S. and Vega, M. and Molina, R. and Katsaggelos, A.K.},
doi = {10.1016/j.dsp.2014.05.017},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Bayesian methods,Parameter estimation,Super-resolution,Total variation,Variational methods},
month = {sep},
pages = {1--10},
title = {{A non-stationary image prior combination in super-resolution}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200414001882},
volume = {32},
year = {2014}
}

A new Bayesian Super-Resolution (SR) image registration and reconstruction method is proposed. The new method utilizes a prior distribution based on a general combination of spatially adaptive, or non-stationary, image filters, which includes an adaptive local strength parameter able to preserve both image edges and textures. With the application of variational techniques, the proposed method allows for the automatic estimation of all problem unknowns. An experimental comparison between state of the art methods and the proposed SR approach has been performed on both synthetic and real images. © 2014 Elsevier Inc.

@article{Martin2014,
abstract = {The ability to accurately estimate effective connectivity among brain regions from neuroimaging data could help answering many open questions in neuroscience. We propose a method which uses causality to obtain a measure of effective connectivity from fMRI data. The method uses a vector autoregressive model for the latent variables describing neuronal activity in combination with a linear observation model based on a convolution with a hemodynamic response function. Due to the employed modeling, it is possible to efficiently estimate all latent variables of the model using a variational Bayesian inference algorithm. The computational efficiency of the method enables us to apply it to large scale problems with high sampling rates and several hundred regions of interest. We use a comprehensive empirical evaluation with synthetic and real fMRI data to evaluate the performance of our method under various conditions. {\textcopyright} 2014 Luessi, Babacan, Molina, Booth and Katsaggelos.},
author = {Luessi, Martin and Babacan, S. Derin and Molina, Rafael and Booth, James R. and Katsaggelos, Aggelos K.},
doi = {10.3389/fninf.2014.00045},
issn = {1662-5196},
journal = {Frontiers in Neuroinformatics},
keywords = {Causality,Connectivity,Granger causality,Variational Bayesian method,fMRI},
month = {may},
number = {MAY},
pages = {45 , publisher = Frontiers Media SA},
title = {{Variational Bayesian causal connectivity analysis for fMRI}},
url = {http://journal.frontiersin.org/article/10.3389/fninf.2014.00045/abstract},
volume = {8},
year = {2014}
}

The ability to accurately estimate effective connectivity among brain regions from neuroimaging data could help answering many open questions in neuroscience. We propose a method which uses causality to obtain a measure of effective connectivity from fMRI data. The method uses a vector autoregressive model for the latent variables describing neuronal activity in combination with a linear observation model based on a convolution with a hemodynamic response function. Due to the employed modeling, it is possible to efficiently estimate all latent variables of the model using a variational Bayesian inference algorithm. The computational efficiency of the method enables us to apply it to large scale problems with high sampling rates and several hundred regions of interest. We use a comprehensive empirical evaluation with synthetic and real fMRI data to evaluate the performance of our method under various conditions. © 2014 Luessi, Babacan, Molina, Booth and Katsaggelos.

@article{Pablo2014b,
author = {Ruiz, Pablo and Madero-Orozco, Hiram and Mateos, Javier and {Osiris Vergara-Villegas}, Osslan and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.sigpro.2013.09.027},
issn = {01651684},
journal = {Signal Processing},
month = {oct},
pages = {296--308},
title = {{Combining Poisson singular integral and total variation prior models in image restoration}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0165168413003927},
volume = {103},
year = {2014}
}

@article{Zhaofu2014a,
abstract = {We propose a Bayesian based algorithm to recover sparse signals from compressed noisy measurements in the presence of a smooth background component. This problem is closely related to robust principal component analysis and compressive sensing, and is found in a number of practical areas. The proposed algorithm adopts a hierarchical Bayesian framework for modeling, and employs approximate inference to estimate the unknowns. Numerical examples demonstrate the effectiveness of the proposed algorithm and its advantage over the current state-of-the-art solutions. {\textcopyright} 1994-2012 IEEE.},
author = {Chen, Zhaofu and Molina, Rafael and Katsaggelos, Aggelos K.},
chapter = {1012},
doi = {10.1109/LSP.2014.2321256},
isbn = {1070-9908 1558-2361},
issn = {1070-9908},
journal = {IEEE Signal Processing Letters},
keywords = {Bayesian algorithm,compressive sensing,robust principal component analysis},
month = {aug},
number = {8},
pages = {1012--1016},
title = {{Automated Recovery of Compressedly Observed Sparse Signals From Smooth Background}},
url = {http://ieeexplore.ieee.org/document/6808512/},
volume = {21},
year = {2014}
}

We propose a Bayesian based algorithm to recover sparse signals from compressed noisy measurements in the presence of a smooth background component. This problem is closely related to robust principal component analysis and compressive sensing, and is found in a number of practical areas. The proposed algorithm adopts a hierarchical Bayesian framework for modeling, and employs approximate inference to estimate the unknowns. Numerical examples demonstrate the effectiveness of the proposed algorithm and its advantage over the current state-of-the-art solutions. © 1994-2012 IEEE.

@article{Ilya2014,
abstract = {Automatic camera calibration has remained a hard topic in computer vision since its inception due to its reliance on the image correspondence problem. This problem becomes even more pronounced when calibrating a depth image with a color image due to a lack of simple correspondences between the two modalities. In this work, we develop a completely automatic, very fast, online algorithm that demonstrates how a consumer-grade depth camera can be calibrated with a color camera with minimal user interaction. {\textcopyright} 2013 Elsevier Inc. All rights reserved.},
author = {Mikhelson, Ilya V. and Lee, Philip G. and Sahakian, Alan V. and Wu, Ying and Katsaggelos, Aggelos K.},
doi = {10.1016/j.jvcir.2013.03.010},
issn = {10473203},
journal = {Journal of Visual Communication and Image Representation},
keywords = {Automatic,Calibration,Color cameras,Depth cameras,Fast,Online,Point cloud,Point correspondence},
month = {jan},
number = {1},
pages = {218--226},
title = {{Automatic, fast, online calibration between depth and color cameras}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1047320313000503},
volume = {25},
year = {2014}
}

Automatic camera calibration has remained a hard topic in computer vision since its inception due to its reliance on the image correspondence problem. This problem becomes even more pronounced when calibrating a depth image with a color image due to a lack of simple correspondences between the two modalities. In this work, we develop a completely automatic, very fast, online algorithm that demonstrates how a consumer-grade depth camera can be calibrated with a color camera with minimal user interaction. © 2013 Elsevier Inc. All rights reserved.

@article{Evaggelia2014,
abstract = {Despite the important properties of unit norm tight frames (UNTFs) and equiangular tight frames (ETFs), their construction has been proven extremely difficult. The few known techniques produce only a small number of such frames while imposing certain restrictions on frame dimensions. Motivated by the application of incoherent tight frames in compressed sensing (CS), we propose a methodology to construct incoherent UNTFs. When frame redundancy is not very high, the achieved maximal column correlation becomes close to the lowest possible bound. The proposed methodology may construct frames of any dimensions. The obtained frames are employed in CS to produce optimized projection matrices. Experimental results show that the proposed optimization technique improves CS signal recovery, increasing the reconstruction accuracy. Considering that the UNTFs and ETFs are important in sparse representations, channel coding, and communications, we expect that the proposed construction will be useful in other applications, besides the CS. {\textcopyright} 1963-2012 IEEE.},
author = {Tsiligianni, Evaggelia V. and Kondi, Lisimachos P. and Katsaggelos, Aggelos K.},
doi = {10.1109/TIT.2014.2308171},
issn = {0018-9448},
journal = {IEEE Transactions on Information Theory},
keywords = {Grassmannian frames,Unit norm tight frames,compressed sensing},
month = {apr},
number = {4},
pages = {2319--2330},
title = {{Construction of Incoherent Unit Norm Tight Frames With Application to Compressed Sensing}},
url = {http://ieeexplore.ieee.org/document/6748087/},
volume = {60},
year = {2014}
}

Despite the important properties of unit norm tight frames (UNTFs) and equiangular tight frames (ETFs), their construction has been proven extremely difficult. The few known techniques produce only a small number of such frames while imposing certain restrictions on frame dimensions. Motivated by the application of incoherent tight frames in compressed sensing (CS), we propose a methodology to construct incoherent UNTFs. When frame redundancy is not very high, the achieved maximal column correlation becomes close to the lowest possible bound. The proposed methodology may construct frames of any dimensions. The obtained frames are employed in CS to produce optimized projection matrices. Experimental results show that the proposed optimization technique improves CS signal recovery, increasing the reconstruction accuracy. Considering that the UNTFs and ETFs are important in sparse representations, channel coding, and communications, we expect that the proposed construction will be useful in other applications, besides the CS. © 1963-2012 IEEE.

@article{Cheuk2014,
author = {Chan, Cheuk L. and Katsaggelos, A. K. and Sahakian, A. V.},
isbn = {9781315078274},
journal = {Medical Imaging Syst Tech and Ap: Cardiovascular Systems},
pages = {93--146},
title = {{Techniques in image sequence filtering for clinical angiography}},
year = {2014}
}

@article{Sotirios2014,
abstract = {This paper describes the use of a customized algorithm for the colorization of historical black and white photographs documenting earlier states of paintings. This study specifically focuses on Pablo Picasso's mid-century Mediterranean masterpiece La Joie de Vivre, 1946 (Mus{\'{e}}e Picasso, Antibes, France). The custom-designed algorithm allows computer-controlled spreading of color information on a digital image of black and white historical photographs to obtain accurate color renditions. Expert observation of the present state of the painting, coupled with stratigraphic information from cross sections allows the attribution of color information to selected pixels in the digitized images. The algorithm uses the localized color information and the grayscale intensities of the black and white historical photographs to formulate a set of equations for the missing color values of the remaining pixels. The computational resolution of such equations allows an accurate colorization that preserves brushwork and shading. This new method is proposed as a valuable alternative to the use of commercial software to apply flat areas of color, which is currently the most common practice for colorization efforts in the conservation community. Availability of such colorized images enhances the art-historical understanding of the works and might lead to better-informed treatment. {\textcopyright} The International Institute for Conservation of Historic and Artistic Works 2014.},
author = {Tsaftaris, Sotirios A. and Casadio, Francesca and Andral, Jean-Louis and Katsaggelos, Aggelos K.},
doi = {10.1179/2047058413Y.0000000104},
issn = {0039-3630},
journal = {Studies in Conservation},
keywords = {Colorization algorithm,Cultural heritage,Historical photographs,Image processing,Pablo Picasso},
month = {may},
number = {3},
pages = {125--135},
title = {{A novel visualization tool for art history and conservation: Automated colorization of black and white archival photographs of works of art}},
url = {http://www.tandfonline.com/doi/full/10.1179/2047058413Y.0000000104},
volume = {59},
year = {2014}
}

This paper describes the use of a customized algorithm for the colorization of historical black and white photographs documenting earlier states of paintings. This study specifically focuses on Pablo Picasso's mid-century Mediterranean masterpiece La Joie de Vivre, 1946 (Musée Picasso, Antibes, France). The custom-designed algorithm allows computer-controlled spreading of color information on a digital image of black and white historical photographs to obtain accurate color renditions. Expert observation of the present state of the painting, coupled with stratigraphic information from cross sections allows the attribution of color information to selected pixels in the digitized images. The algorithm uses the localized color information and the grayscale intensities of the black and white historical photographs to formulate a set of equations for the missing color values of the remaining pixels. The computational resolution of such equations allows an accurate colorization that preserves brushwork and shading. This new method is proposed as a valuable alternative to the use of commercial software to apply flat areas of color, which is currently the most common practice for colorization efforts in the conservation community. Availability of such colorized images enhances the art-historical understanding of the works and might lead to better-informed treatment. © The International Institute for Conservation of Historic and Artistic Works 2014.

@article{Tao2013,
abstract = {Compressive sensing (CS) is an innovative technology, allowing us to capture signals with significantly fewer samples than those required by classical Nyquist theory. We propose a novel adaptive video compressive sensing algorithm to exploit the potential of CS in video acquisition. Each frame is divided into blocks to take advantage of its inhomogeneity. We first classify the blocks into one of three types based on their texture complexity and their temporal difference from neighboring frames based on which we determine the number of required measurements. In the reconstruction process, we use the measurements made for the later frames to assist the recovery of previous ones, thus ensuring improved reconstruction quality even when the number of measurements for each frame is limited. Our experimental results demonstrate that we not only obtain significant visual quality improvement but also achieve at least 2.5 dB gain in peak signal- to-noise ratio compared with the existing video compressive sensing algorithms. {\textcopyright} SPIE and IS&T.},
author = {Li, Tao and Wang, Xiaohua and Wang, Weihe and Katsaggelos, Aggelos K.},
doi = {10.1117/1.JEI.22.4.043003},
issn = {1017-9909},
journal = {Journal of Electronic Imaging},
month = {oct},
number = {4},
pages = {043003},
title = {{Compressive video sensing with limited measurements}},
url = {http://electronicimaging.spiedigitallibrary.org/article.aspx?doi=10.1117/1.JEI.22.4.043003},
volume = {22},
year = {2013}
}

Compressive sensing (CS) is an innovative technology, allowing us to capture signals with significantly fewer samples than those required by classical Nyquist theory. We propose a novel adaptive video compressive sensing algorithm to exploit the potential of CS in video acquisition. Each frame is divided into blocks to take advantage of its inhomogeneity. We first classify the blocks into one of three types based on their texture complexity and their temporal difference from neighboring frames based on which we determine the number of required measurements. In the reconstruction process, we use the measurements made for the later frames to assist the recovery of previous ones, thus ensuring improved reconstruction quality even when the number of measurements for each frame is limited. Our experimental results demonstrate that we not only obtain significant visual quality improvement but also achieve at least 2.5 dB gain in peak signal- to-noise ratio compared with the existing video compressive sensing algorithms. © SPIE and IS&T.

@article{xin2013laplacian,
abstract = {Visual query-by-capture applications call for a compact visual descriptor with minimum descriptor length. Preserving the visual identification performance while minimising the bit rate is a focus of the on-going MPEG7 CDVS (Compact Descriptors for Visual Search) standardisation effort. In this paper we tackle this problem by adopting Laplacian embedding for SIFT feature compression and employing topology verification based on a novel graph cut measure. In contrast to previous feature compression schemes, we approach the problem by finding a Laplacian embedding that preserves the nearest neighbour relations in feature space. Furthermore, we develop an efficient yet effective topology verification (TV) scheme to perform spatial consistency checking. In contrast to previous works on geometric verification, instead of enumerating all possible combinations of coordinate alignments of an image pair, this TV solution verifies possibly misaligned coordinate sets with a learning method which acquires a proper boundary between the topology representation of matched and non-matched image pairs. Furthermore, this TV solution is invariant to in-plane rotation, scaling and is quite resilient to a range of out-of-plane rotations. The proposed Laplacian embedding and Topological verification scheme are tested with the CDVS dataset and are found to be effective. {\textcopyright} 2012 Elsevier B.V.},
author = {Xin, Xin and Li, Zhu and Katsaggelos, Aggelos K.},
doi = {10.1016/j.image.2012.11.003},
issn = {09235965},
journal = {Signal Processing: Image Communication},
keywords = {Geometrical re-ranking,Laplacian embedding,Mobile visual search,Point set topology,Visual identification},
month = {apr},
number = {4},
pages = {323--333},
publisher = {Elsevier},
title = {{Laplacian embedding and key points topology verification for large scale mobile visual identification}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0923596512002056},
volume = {28},
year = {2013}
}

Visual query-by-capture applications call for a compact visual descriptor with minimum descriptor length. Preserving the visual identification performance while minimising the bit rate is a focus of the on-going MPEG7 CDVS (Compact Descriptors for Visual Search) standardisation effort. In this paper we tackle this problem by adopting Laplacian embedding for SIFT feature compression and employing topology verification based on a novel graph cut measure. In contrast to previous feature compression schemes, we approach the problem by finding a Laplacian embedding that preserves the nearest neighbour relations in feature space. Furthermore, we develop an efficient yet effective topology verification (TV) scheme to perform spatial consistency checking. In contrast to previous works on geometric verification, instead of enumerating all possible combinations of coordinate alignments of an image pair, this TV solution verifies possibly misaligned coordinate sets with a learning method which acquires a proper boundary between the topology representation of matched and non-matched image pairs. Furthermore, this TV solution is invariant to in-plane rotation, scaling and is quite resilient to a range of out-of-plane rotations. The proposed Laplacian embedding and Topological verification scheme are tested with the CDVS dataset and are found to be effective. © 2012 Elsevier B.V.

@article{Martin2013,
abstract = {In the simultaneous sparse approximation problem, several latent vectors corresponding to independent random realizations from a common sparsity profile are recovered from an undercomplete set of measurements. In this paper, we address an extension of this problem, where in addition to the common sparsity profile, the vectors of interest are assumed to have a high correlation among each other. Specifically, we consider the case when the non-zero rows in the combined latent signal vectors are considered to be temporally smooth signals. We present a Bayesian formulation of the problem and develop a greedy inference algorithmbased on sparse Bayesian learning for independent observations. A difficulty is that unlike for existing greedy methods, there is no closed form expression for the maximizer of the objective function in the greedy algorithm when row correlations are introduced. We derive two methods to maximize the objective function, one based on the EM algorithm and another on a fixed-point iteration. Empirical results show that the proposed method provides better reconstruction results compared to existing methods, especially when the signal-to-noise ratio is low and the latent signal vectors are highly correlated. We also demonstrate the application of the proposed method to source localization in magnetoencephalography, where it obtains a temporally smooth solution with accurate localization of the brain activity. {\textcopyright} 1991-2012 IEEE.},
author = {Luessi, Martin and Babacan, S. Derin and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TSP.2013.2280441},
issn = {1053-587X},
journal = {IEEE Transactions on Signal Processing},
keywords = {Bayesian methods,MEG/EEG source localization,Sparse Bayesian learning,Sparse signal recovery,Sparsity},
month = {nov},
number = {22},
pages = {5716--5729},
title = {{Bayesian Simultaneous Sparse Approximation With Smooth Signals}},
url = {http://ieeexplore.ieee.org/document/6588611/},
volume = {61},
year = {2013}
}

In the simultaneous sparse approximation problem, several latent vectors corresponding to independent random realizations from a common sparsity profile are recovered from an undercomplete set of measurements. In this paper, we address an extension of this problem, where in addition to the common sparsity profile, the vectors of interest are assumed to have a high correlation among each other. Specifically, we consider the case when the non-zero rows in the combined latent signal vectors are considered to be temporally smooth signals. We present a Bayesian formulation of the problem and develop a greedy inference algorithmbased on sparse Bayesian learning for independent observations. A difficulty is that unlike for existing greedy methods, there is no closed form expression for the maximizer of the objective function in the greedy algorithm when row correlations are introduced. We derive two methods to maximize the objective function, one based on the EM algorithm and another on a fixed-point iteration. Empirical results show that the proposed method provides better reconstruction results compared to existing methods, especially when the signal-to-noise ratio is low and the latent signal vectors are highly correlated. We also demonstrate the application of the proposed method to source localization in magnetoencephalography, where it obtains a temporally smooth solution with accurate localization of the brain activity. © 1991-2012 IEEE.

@article{ye2013wireless,
abstract = {A wireless video surveillance system consists of three major components: 1) the video capture and preprocessing; 2) the video compression and transmission in wireless sensor networks; and 3) the video analysis at the receiving end. A myriad of research works have been dedicated to this field due to its increasing popularity in surveillance applications. This survey provides a comprehensive overview of existing state-of-the-Art technologies developed for wireless video surveillance, based on the in-depth analysis of the requirements and challenges in current systems. Specifically, the physical network infrastructure for video transmission over wireless channel is analyzed. The representative technologies for video capture and preliminary vision tasks are summarized. For video compression and transmission over the wireless networks, the ultimate goal is to maximize the received video quality under the resource limitation. This is also the main focus of this survey. We classify different schemes into categories including unequal error protection, error resilience, scalable video coding, distributed video coding, and cross-layer control. Crosslayer control proves to be a desirable measure for system-level optimal resource allocation. At the receiver's end, the received video is further processed for higher-level vision tasks, and the security and privacy issues in surveillance applications are also discussed. {\textcopyright} 2013 IEEE.},
author = {Ye, Yun and Ci, Song and Katsaggelos, Aggelos K. and Liu, Yanwei and Qian, Yi and {Yun Ye} and {Song Ci} and Katsaggelos, Aggelos K. and {Yanwei Liu} and {Yi Qian}},
doi = {10.1109/ACCESS.2013.2282613},
issn = {2169-3536},
journal = {IEEE Access},
keywords = {Cross-layer control,Multimedia communications,Video analysis,Video surveillance,Wireless sensor networks},
pages = {646--660},
publisher = {IEEE},
title = {{Wireless Video Surveillance: A Survey}},
url = {http://ieeexplore.ieee.org/document/6603291/},
volume = {1},
year = {2013}
}

A wireless video surveillance system consists of three major components: 1) the video capture and preprocessing; 2) the video compression and transmission in wireless sensor networks; and 3) the video analysis at the receiving end. A myriad of research works have been dedicated to this field due to its increasing popularity in surveillance applications. This survey provides a comprehensive overview of existing state-of-the-Art technologies developed for wireless video surveillance, based on the in-depth analysis of the requirements and challenges in current systems. Specifically, the physical network infrastructure for video transmission over wireless channel is analyzed. The representative technologies for video capture and preliminary vision tasks are summarized. For video compression and transmission over the wireless networks, the ultimate goal is to maximize the received video quality under the resource limitation. This is also the main focus of this survey. We classify different schemes into categories including unequal error protection, error resilience, scalable video coding, distributed video coding, and cross-layer control. Crosslayer control proves to be a desirable measure for system-level optimal resource allocation. At the receiver's end, the received video is further processed for higher-level vision tasks, and the security and privacy issues in surveillance applications are also discussed. © 2013 IEEE.

@article{Changhun2013,
author = {Cho, Changhun and Katsaggelos, Aggelos K and Paik, Joonki},
journal = {IEIE Transactions on Smart Processing and Computing},
keywords = {fast discrete curvelet transform,fdct,image restoration,vaguelette-curvelet decomposition},
number = {3},
pages = {140--147},
title = {{The Vaguelette-Curvelet Decomposition for Image Deblurring}},
volume = {2},
year = {2013}
}

@article{Lu2013,
abstract = {In general, subpixel-based downsampling can achieve higher apparent resolution of the down-sampled images on LCD or OLED displays than pixel-based downsampling. With the frequency domain analysis of subpixel-based downsampling, we discover special characteristics of the luma-chroma color transform choice for monochrome images. With these, we model the anti-aliasing filter design for subpixel-based monochrome image downsampling as a human visual system-based optimization problem with a two-term cost function and obtain a closed-form solution. One cost term measures the luminance distortion and the other term measures the chrominance aliasing in our chosen luma-chroma space. Simulation results suggest that the proposed method can achieve sharper down-sampled gray/font images compared with conventional pixel and subpixel-based methods, without noticeable color fringing artifacts. {\textcopyright} 1992-2012 IEEE.},
author = {Fang, Lu and Au, Oscar C. and Cheung, Ngai-Man and Katsaggelos, Aggelos K. and Li, Houqiang and Zou, Feng},
doi = {10.1109/TIP.2013.2262288},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Subpixel rendering,down-sampling,human visual system,improve resolution},
month = {oct},
number = {10},
pages = {3818--3829},
title = {{Luma-Chroma Space Filter Design for Subpixel-Based Monochrome Image Downsampling}},
url = {http://ieeexplore.ieee.org/document/6514934/},
volume = {22},
year = {2013}
}

In general, subpixel-based downsampling can achieve higher apparent resolution of the down-sampled images on LCD or OLED displays than pixel-based downsampling. With the frequency domain analysis of subpixel-based downsampling, we discover special characteristics of the luma-chroma color transform choice for monochrome images. With these, we model the anti-aliasing filter design for subpixel-based monochrome image downsampling as a human visual system-based optimization problem with a two-term cost function and obtain a closed-form solution. One cost term measures the luminance distortion and the other term measures the chrominance aliasing in our chosen luma-chroma space. Simulation results suggest that the proposed method can achieve sharper down-sampled gray/font images compared with conventional pixel and subpixel-based methods, without noticeable color fringing artifacts. © 1992-2012 IEEE.

@article{Amizic2013,
author = {Amizic, Bruno and Spinoulas, Leonidas and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2013.2266100},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {oct},
number = {10},
pages = {3994--4006},
title = {{Compressive Blind Image Deconvolution}},
url = {https://ieeexplore.ieee.org/document/6523098/},
volume = {22},
year = {2013}
}

@article{xin2013large,
abstract = {Visual search over large image repositories in real time is one of the key challenges for applications such as mobile visual query-by-capture, augmented reality, and biometrics-based identification. Search accuracy and response speed are two important performance factors. This article focuses on one of the important elements of this technology that enables large-scale visual search: indexing (or hashing). Indexing is the process of organizing a database of searchable elements into an efficiently searchable configuration. The searchable elements in our case are compact features extracted from images. This article explores a new indexing scheme. The authors optimize the design of a hash-code collision and counting scheme to enable fast search of visual features of MPEG CDVS. {\textcopyright} 1994-2012 IEEE.},
author = {Xin, Xin and Nagar, Abhishek and Srivastava, Gaurav and Li, Zhu and Fernandes, Felix and Katsaggelos, Aggelos K.},
doi = {10.1109/MMUL.2013.22},
issn = {1070-986X},
journal = {IEEE MultiMedia},
keywords = {CDVS,MPEG,hash-code collision,indexing,large-scale visual search,mobile search,multimedia,multimedia applications},
month = {apr},
number = {2},
pages = {62--71},
publisher = {IEEE},
title = {{Large Visual Repository Search with Hash Collision Design Optimization}},
url = {http://ieeexplore.ieee.org/document/6530593/},
volume = {20},
year = {2013}
}

Visual search over large image repositories in real time is one of the key challenges for applications such as mobile visual query-by-capture, augmented reality, and biometrics-based identification. Search accuracy and response speed are two important performance factors. This article focuses on one of the important elements of this technology that enables large-scale visual search: indexing (or hashing). Indexing is the process of organizing a database of searchable elements into an efficiently searchable configuration. The searchable elements in our case are compact features extracted from images. This article explores a new indexing scheme. The authors optimize the design of a hash-code collision and counting scheme to enable fast search of visual features of MPEG CDVS. © 1994-2012 IEEE.

@article{Salvador2013a,
abstract = {In this paper the application of image prior combinations to the Bayesian Super Resolution (SR) image registration and reconstruction problem is studied. Two sparse image priors, a Total Variation (TV) prior and a prior based on the ℓ1 norm of horizontal and vertical first-order differences (f.o.d.), are combined with a non-sparse Simultaneous Auto Regressive (SAR) prior. Since, for a given observation model, each prior produces a different posterior distribution of the underlying High Resolution (HR) image, the use of variational approximation will produce as many posterior approximations as priors we want to combine. A unique approximation is obtained here by finding the distribution on the HR image given the observations that minimizes a linear convex combination of Kullback-Leibler (KL) divergences. We find this distribution in closed form. The estimated HR images are compared with the ones obtained by other SR reconstruction methods. {\textcopyright} 2012 Elsevier Inc.},
author = {Villena, S. and Vega, M. and Babacan, S.D. and Molina, R. and Katsaggelos, A.K.},
doi = {10.1016/j.dsp.2012.10.002},
issn = {10512004},
journal = {Digital Signal Processing},
keywords = {Bayesian methods,Parameter estimation,Super resolution,Total variation,Variational methods},
month = {mar},
number = {2},
pages = {530--541},
title = {{Bayesian combination of sparse and non-sparse priors in image super resolution}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1051200412002400},
volume = {23},
year = {2013}
}

In this paper the application of image prior combinations to the Bayesian Super Resolution (SR) image registration and reconstruction problem is studied. Two sparse image priors, a Total Variation (TV) prior and a prior based on the ℓ1 norm of horizontal and vertical first-order differences (f.o.d.), are combined with a non-sparse Simultaneous Auto Regressive (SAR) prior. Since, for a given observation model, each prior produces a different posterior distribution of the underlying High Resolution (HR) image, the use of variational approximation will produce as many posterior approximations as priors we want to combine. A unique approximation is obtained here by finding the distribution on the HR image given the observations that minimizes a linear convex combination of Kullback-Leibler (KL) divergences. We find this distribution in closed form. The estimated HR images are compared with the ones obtained by other SR reconstruction methods. © 2012 Elsevier Inc.

@article{chen2013application,
abstract = {In this paper, we present a transportation video coding and wireless transmission system specifically tailored to automated vehicle tracking applications. By taking into account the video characteristics and the lossy nature of the wireless channels, we propose video preprocessing and error control approaches to enhance tracking performance while conserving bandwidth resources and computational power at the transmitter. Compared with current state-of-the-art H.264-based implementations, our system is shown to yield over 80% bitrate savings for comparable tracking accuracy. {\textcopyright} 2000-2011 IEEE.},
author = {Chen, Zhaofu and Tsaftaris, Sotirios A. and Soyak, Eren and Katsaggelos, Aggelos K.},
doi = {10.1109/TITS.2013.2266861},
issn = {1524-9050},
journal = {IEEE Transactions on Intelligent Transportation Systems},
keywords = {Error concealment (ERC),Forward error correction (FEC),H.264/AVC,Object tracking,Preprocessing,Surveillance centric coding,Transportation video},
month = {dec},
number = {4},
pages = {2002--2007},
publisher = {IEEE},
title = {{Application-Aware Approach to Compression and Transmission of H.264 Encoded Video for Automated and Centralized Transportation Surveillance}},
url = {https://ieeexplore.ieee.org/document/6544640},
volume = {14},
year = {2013}
}

In this paper, we present a transportation video coding and wireless transmission system specifically tailored to automated vehicle tracking applications. By taking into account the video characteristics and the lossy nature of the wireless channels, we propose video preprocessing and error control approaches to enhance tracking performance while conserving bandwidth resources and computational power at the transmitter. Compared with current state-of-the-art H.264-based implementations, our system is shown to yield over 80% bitrate savings for comparable tracking accuracy. © 2000-2011 IEEE.

@article{Miguel2013,
abstract = {In this paper we propose a space-variant blur estimation and effective denoising/deconvolution method for combining a long exposure blurry image with a short exposure noisy one. The blur in the long exposure shot is mainly caused by camera shake or object motion, and the noise in the underexposed image is introduced by the gain factor applied to the sensor when the ISO is set to an high value. Due to the space variant degradation, the image pair is divided into overlapping patches for processing. The main idea in the deconvolution algorithm is to incorporate a combination of prior image models into a spatially-varying deblurring/denoising framework which is applied to each patch. The method employs a kernel and parameter estimation method to choose between denoising or deblurring each patch. Experiments on both synthetic and real images are provided to validate the proposed approach. {\textcopyright} 2012 Elsevier B.V.},
author = {Tall{\'{o}}n, Miguel and Mateos, Javier and Babacan, S. Derin and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.inffus.2012.08.003},
issn = {15662535},
journal = {Information Fusion},
keywords = {Blind deconvolution,Image denoising,Image fusion,Low light imaging,Motion blur},
month = {oct},
number = {4},
pages = {396--409},
title = {{Space-variant blur deconvolution and denoising in the dual exposure problem}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S156625351200067X},
volume = {14},
year = {2013}
}

In this paper we propose a space-variant blur estimation and effective denoising/deconvolution method for combining a long exposure blurry image with a short exposure noisy one. The blur in the long exposure shot is mainly caused by camera shake or object motion, and the noise in the underexposed image is introduced by the gain factor applied to the sensor when the ISO is set to an high value. Due to the space variant degradation, the image pair is divided into overlapping patches for processing. The main idea in the deconvolution algorithm is to incorporate a combination of prior image models into a spatially-varying deblurring/denoising framework which is applied to each patch. The method employs a kernel and parameter estimation method to choose between denoising or deblurring each patch. Experiments on both synthetic and real images are provided to validate the proposed approach. © 2012 Elsevier B.V.

@article{Esteban2013,
abstract = {In this paper, we propose an algorithm for image restoration based on fusing nonstationary edgepreserving priors. We develop a Bayesian modeling followed by an evidence approximation inference approach for deriving the analytic foundations of the proposed restoration method. Through a series of approximations, the final implementation of the proposed image restoration algorithm is iterative and takes advantage of the Fourier domain. Simulation results over a variety of blurred and noisy standard test images indicate that the presented method comfortably surpasses the current state-of-the-art image restoration for compactly supported degradations. We finally present experimental results by digitally refocusing images captured with controlled defocus, successfully confirming the ability of the proposed restoration algorithm in recovering extra features and rich details, while still preserving edges. {\textcopyright} 2013 Optical Society of America.},
author = {Vera, Esteban and Vega, Miguel and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1364/AO.52.00D102},
issn = {1559-128X},
journal = {Applied Optics},
month = {apr},
number = {10},
pages = {D102},
title = {{Iterative image restoration using nonstationary priors}},
url = {https://opg.optica.org/abstract.cfm?URI=ao-52-10-D102},
volume = {52},
year = {2013}
}

In this paper, we propose an algorithm for image restoration based on fusing nonstationary edgepreserving priors. We develop a Bayesian modeling followed by an evidence approximation inference approach for deriving the analytic foundations of the proposed restoration method. Through a series of approximations, the final implementation of the proposed image restoration algorithm is iterative and takes advantage of the Fourier domain. Simulation results over a variety of blurred and noisy standard test images indicate that the presented method comfortably surpasses the current state-of-the-art image restoration for compactly supported degradations. We finally present experimental results by digitally refocusing images captured with controlled defocus, successfully confirming the ability of the proposed restoration algorithm in recovering extra features and rich details, while still preserving edges. © 2013 Optical Society of America.

@article{Yuan2012,
abstract = {Given a collection of images or a short video sequence, we define a thematic object as the key object that frequently appears and is the representative of the visual contents. Successful discovery of the thematic object is helpful for object search and tagging, video summarization and understanding, etc. However, this task is challenging because 1) there lacks a priori knowledge of the thematic objects, such as their shapes, scales, locations, and times of re-occurrences, and 2) the thematic object of interest can be under severe variations in appearances due to viewpoint and lighting condition changes, scale variations, etc. Instead of using a top-down generative model to discover thematic visual patterns, we propose a novel bottom-up approach to gradually prune uncommon local visual primitives and recover the thematic objects. A multilayer candidate pruning procedure is designed to accelerate the image data mining process. Our solution can efficiently locate thematic objects of various sizes and can tolerate large appearance variations of the same thematic object. Experiments on challenging image and video data sets and comparisons with existing methods validate the effectiveness of our method. {\textcopyright} 2011 IEEE.},
author = {Yuan, Junsong and Zhao, Gangqiang and Fu, Yun and Li, Zhu and Katsaggelos, Aggelos K. and Wu, Ying and {Junsong Yuan} and {Gangqiang Zhao} and {Yun Fu} and {Zhu Li} and Katsaggelos, Aggelos K. and {Ying Wu}},
doi = {10.1109/TIP.2011.2181952},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Image data mining,thematic object discovery},
month = {apr},
number = {4},
pages = {2207--2219},
pmid = {22207639},
title = {{Discovering Thematic Objects in Image Collections and Videos}},
url = {http://ieeexplore.ieee.org/document/6112717/},
volume = {21},
year = {2012}
}

Given a collection of images or a short video sequence, we define a thematic object as the key object that frequently appears and is the representative of the visual contents. Successful discovery of the thematic object is helpful for object search and tagging, video summarization and understanding, etc. However, this task is challenging because 1) there lacks a priori knowledge of the thematic objects, such as their shapes, scales, locations, and times of re-occurrences, and 2) the thematic object of interest can be under severe variations in appearances due to viewpoint and lighting condition changes, scale variations, etc. Instead of using a top-down generative model to discover thematic visual patterns, we propose a novel bottom-up approach to gradually prune uncommon local visual primitives and recover the thematic objects. A multilayer candidate pruning procedure is designed to accelerate the image data mining process. Our solution can efficiently locate thematic objects of various sizes and can tolerate large appearance variations of the same thematic object. Experiments on challenging image and video data sets and comparisons with existing methods validate the effectiveness of our method. © 2011 IEEE.

@article{Ehsan2012,
abstract = {We consider the problem of foresighted multimedia resource reciprocation in peer-to-peer (P2P) networks, which consist of rational peers aiming at maximizing their individual utilities. We introduce an artificial currency (credit) to take into account the characteristics of different parts of the video signal. The resource reciprocation with the proposed credit metric can be formulated as a stochastic game, in which the peers determine their optimal strategies using Markov Decision Process (MDP) framework. The introduced framework can be applied to the general video coding, and in particular, is suitable for the scalable video where various parts of the encoded bit stream have significantly different importance for the video quality. {\textcopyright} 2012 Elsevier B.V. All rights reserved.},
author = {Maani, Ehsan and Chen, Zhaofu and Katsaggelos, Aggelos K.},
doi = {10.1016/j.image.2012.02.015},
issn = {09235965},
journal = {Signal Processing: Image Communication},
keywords = {Game theory,P2P multimedia sharing},
month = {may},
number = {5},
pages = {545--554},
title = {{A game theoretic approach to video streaming over peer-to-peer networks}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0923596512000446},
volume = {27},
year = {2012}
}

We consider the problem of foresighted multimedia resource reciprocation in peer-to-peer (P2P) networks, which consist of rational peers aiming at maximizing their individual utilities. We introduce an artificial currency (credit) to take into account the characteristics of different parts of the video signal. The resource reciprocation with the proposed credit metric can be formulated as a stochastic game, in which the peers determine their optimal strategies using Markov Decision Process (MDP) framework. The introduced framework can be applied to the general video coding, and in particular, is suitable for the scalable video where various parts of the encoded bit stream have significantly different importance for the video quality. © 2012 Elsevier B.V. All rights reserved.

@article{babacan2012sparse,
abstract = {Recovery of low-rank matrices has recently seen significant activity in many areas of science and engineering, motivated by recent theoretical results for exact reconstruction guarantees and interesting practical applications. In this paper, we present novel recovery algorithms for estimating low-rank matrices in matrix completion and robust principal component analysis based on sparse Bayesian learning (SBL) principles. Starting from a matrix factorization formulation and enforcing the low-rank constraint in the estimates as a sparsity constraint, we develop an approach that is very effective in determining the correct rank while providing high recovery performance. We provide connections with existing methods in other similar problems and empirical results and comparisons with current state-of-the-art methods that illustrate the effectiveness of this approach. {\textcopyright} 2012 IEEE.},
archivePrefix = {arXiv},
arxivId = {1102.5288},
author = {Babacan, S. Derin and Luessi, Martin and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TSP.2012.2197748},
eprint = {1102.5288},
issn = {1053-587X},
journal = {IEEE Transactions on Signal Processing},
keywords = {Bayesian methods,low-rankness,matrix completion,outlier detection,robust principal component analysis,sparse Bayesian learning,sparsity,variational Bayesian inference},
month = {aug},
number = {8},
pages = {3964--3977},
publisher = {IEEE},
title = {{Sparse Bayesian Methods for Low-Rank Matrix Estimation}},
url = {http://ieeexplore.ieee.org/document/6194350/},
volume = {60},
year = {2012}
}

Recovery of low-rank matrices has recently seen significant activity in many areas of science and engineering, motivated by recent theoretical results for exact reconstruction guarantees and interesting practical applications. In this paper, we present novel recovery algorithms for estimating low-rank matrices in matrix completion and robust principal component analysis based on sparse Bayesian learning (SBL) principles. Starting from a matrix factorization formulation and enforcing the low-rank constraint in the estimates as a sparsity constraint, we develop an approach that is very effective in determining the correct rank while providing high recovery performance. We provide connections with existing methods in other similar problems and empirical results and comparisons with current state-of-the-art methods that illustrate the effectiveness of this approach. © 2012 IEEE.

@article{Lu2012,
abstract = {A portable device such as a digital camera with a single sensor and Bayer color filter array (CFA) requires demosaicing to reconstruct a full color image. To display a high resolution image on a low resolution LCD screen of the portable device, it must be down-sampled. The two steps, demosaicing and down-sampling, influence each other. On one hand, the color artifacts introduced in demosaicing may be magnified when followed by down-sampling; on the other hand, the detail removed in the down-sampling cannot be recovered in the demosaicing. Therefore, it is very important to consider simultaneous demosaicing and down-sampling. In this paper, we propose a fast frequency-domain analysis approach for joint demosaicing and subpixel-based down-sampling (FFA-JDSD) of single sensor Bayer images. In FFA-JDSD, we integrate demosaicing into down-sampling by directly performing subpixel- based down-sampling in the Bayer domain, due to which the computational complexity is significantly reduced. We use a frequency domain analysis tool to show that the cut-off frequency of the low-pass filter for JDSD can be effectively extended beyond the Nyquist frequency, resulting in much sharper down-sampled images. Experimental results show that, compared with the methods that separately perform demosaicing and down-sampling, FFAJDSD has much less computational complexity, and can produce much sharper results at the expense of slight color fringing artifact. {\textcopyright} 2012 IEEE.},
author = {Fang, Lu and Au, Oscar C. and Chen, Yan and Katsaggelos, Aggelos K. and Wang, Hanli and Wen, Xing},
doi = {10.1109/TMM.2012.2191269},
issn = {1520-9210},
journal = {IEEE Transactions on Multimedia},
keywords = {Demosaicking,down-sampling,subpixel rendering},
month = {aug},
number = {4},
pages = {1359--1369},
title = {{Joint Demosaicing and Subpixel-Based Down-Sampling for Bayer Images: A Fast Frequency-Domain Analysis Approach}},
url = {http://ieeexplore.ieee.org/document/6171859/},
volume = {14},
year = {2012}
}

A portable device such as a digital camera with a single sensor and Bayer color filter array (CFA) requires demosaicing to reconstruct a full color image. To display a high resolution image on a low resolution LCD screen of the portable device, it must be down-sampled. The two steps, demosaicing and down-sampling, influence each other. On one hand, the color artifacts introduced in demosaicing may be magnified when followed by down-sampling; on the other hand, the detail removed in the down-sampling cannot be recovered in the demosaicing. Therefore, it is very important to consider simultaneous demosaicing and down-sampling. In this paper, we propose a fast frequency-domain analysis approach for joint demosaicing and subpixel-based down-sampling (FFA-JDSD) of single sensor Bayer images. In FFA-JDSD, we integrate demosaicing into down-sampling by directly performing subpixel- based down-sampling in the Bayer domain, due to which the computational complexity is significantly reduced. We use a frequency domain analysis tool to show that the cut-off frequency of the low-pass filter for JDSD can be effectively extended beyond the Nyquist frequency, resulting in much sharper down-sampled images. Experimental results show that, compared with the methods that separately perform demosaicing and down-sampling, FFAJDSD has much less computational complexity, and can produce much sharper results at the expense of slight color fringing artifact. © 2012 IEEE.

@article{Miguel2012,
abstract = {In this work we develop a variational framework for the combination of several prior models in Bayesian image restoration and apply it to astronomical images. Since each combination of a given observation model and a prior model produces a different posterior distribution of the underlying image, the use of variational posterior distribution approximation on each posterior will produce as many posterior approximations as priors we want to combine. A unique approximation is obtained here by finding the distribution on the unknown image given the observations that minimizes a linear convex combination of the Kullback-Leibler divergences associated with each posterior distribution. We find this distribution in closed form and also relate the proposed approach to other prior combination methods in the literature. Experimental results on both synthetic images and on real astronomical images validate the proposed approach. {\textcopyright} 2011 Elsevier B.V..},
author = {Vega, Miguel and Mateos, Javier and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1016/j.stamet.2011.04.003},
issn = {15723127},
journal = {Statistical Methodology},
keywords = {Astronomical image processing,Bayesian methods,Model combination,Variational methods},
month = {jan},
number = {1-2},
pages = {19--31},
title = {{Astronomical image restoration using variational methods and model combination}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1572312711000347},
volume = {9},
year = {2012}
}

In this work we develop a variational framework for the combination of several prior models in Bayesian image restoration and apply it to astronomical images. Since each combination of a given observation model and a prior model produces a different posterior distribution of the underlying image, the use of variational posterior distribution approximation on each posterior will produce as many posterior approximations as priors we want to combine. A unique approximation is obtained here by finding the distribution on the unknown image given the observations that minimizes a linear convex combination of the Kullback-Leibler divergences associated with each posterior distribution. We find this distribution in closed form and also relate the proposed approach to other prior combination methods in the literature. Experimental results on both synthetic images and on real astronomical images validate the proposed approach. © 2011 Elsevier B.V..

@article{Ilya2012,
abstract = {This paper presents a solution to an aiming problem in the remote sensing of vital signs using an integration of two systems. The problem is that to collect meaningful data with a millimeter-wave sensor, the antenna must be pointed very precisely at the subjects chest. Even small movements could make the data unreliable. To solve this problem, we attached a camera to the millimeter-wave antenna, and mounted this combined system on a pan/tilt base. Our algorithm initially finds a subjects face and then tracks him/her through subsequent frames, while calculating the position of the subjects chest. For each frame, the camera sends the location of the chest to the pan/tilt base, which rotates accordingly to make the antenna point at the subjects chest. This paper presents a system for concurrent tracking and data acquisition with results from some sample scenarios. {\textcopyright} 2012 IEEE.},
author = {Mikhelson, Ilya V. and Lee, Philip and Bakhtiari, Sasan and Elmer, Thomas W. and Katsaggelos, Aggelos K. and Sahakian, Alan V.},
doi = {10.1109/TITB.2012.2204760},
issn = {1089-7771},
journal = {IEEE Transactions on Information Technology in Biomedicine},
keywords = {Heart rate,human tracking,millimeter wave,patient monitoring,remote sensing},
month = {sep},
number = {5},
pages = {927--934},
pmid = {22711781},
title = {{Noncontact Millimeter-Wave Real-Time Detection and Tracking of Heart Rate on an Ambulatory Subject}},
url = {http://ieeexplore.ieee.org/document/6217314/},
volume = {16},
year = {2012}
}

This paper presents a solution to an aiming problem in the remote sensing of vital signs using an integration of two systems. The problem is that to collect meaningful data with a millimeter-wave sensor, the antenna must be pointed very precisely at the subjects chest. Even small movements could make the data unreliable. To solve this problem, we attached a camera to the millimeter-wave antenna, and mounted this combined system on a pan/tilt base. Our algorithm initially finds a subjects face and then tracks him/her through subsequent frames, while calculating the position of the subjects chest. For each frame, the camera sends the location of the chest to the pan/tilt base, which rotates accordingly to make the antenna point at the subjects chest. This paper presents a system for concurrent tracking and data acquisition with results from some sample scenarios. © 2012 IEEE.

@article{Leonidas2012,
abstract = {In this paper, we briefly describe a single detector passive millimeter-wave imaging system, which has been previously presented. The system uses a cyclic sensing matrix to acquire incoherent measurements of the observed scene and then reconstructs the image using a Bayesian approach. The cyclic nature of the sensing matrix allows for the design of a single unified and compact mask that provides all the required random masks in a convenient way, such that no mechanical mask exchange is needed. Based on this setup, we primarily propose the optimal adaptive selection of sampling submasks out of the full cyclic mask to obtain improved reconstruction results. The reconstructed images show the feasibility of the imaging system as well as its improved performance through the proposed sampling scheme. {\textcopyright} 2012 Optical Society of America.},
author = {Spinoulas, Leonidas and Qi, Jin and Katsaggelos, Aggelos K. and Elmer, Thomas W. and Gopalsami, Nachappa and Raptis, Apostolos C.},
doi = {10.1364/AO.51.006335},
issn = {1559-128X},
journal = {Applied Optics},
month = {sep},
number = {26},
pages = {6335},
title = {{Optimized compressive sampling for passive millimeter-wave imaging}},
url = {https://opg.optica.org/abstract.cfm?URI=ao-51-26-6335},
volume = {51},
year = {2012}
}

In this paper, we briefly describe a single detector passive millimeter-wave imaging system, which has been previously presented. The system uses a cyclic sensing matrix to acquire incoherent measurements of the observed scene and then reconstructs the image using a Bayesian approach. The cyclic nature of the sensing matrix allows for the design of a single unified and compact mask that provides all the required random masks in a convenient way, such that no mechanical mask exchange is needed. Based on this setup, we primarily propose the optimal adaptive selection of sampling submasks out of the full cyclic mask to obtain improved reconstruction results. The reconstructed images show the feasibility of the imaging system as well as its improved performance through the proposed sampling scheme. © 2012 Optical Society of America.

@article{Nachappa2012,
abstract = {Passive millimeter-wave (PMMW) imagers using a single radiometer, called single pixel imagers, employ raster scanning to produce images. A serious drawback of such a single pixel imaging system is the long acquisition time needed to produce a high-fidelity image, arising from two factors: (a) the time to scan the whole scene pixel by pixel and (b) the integration time for each pixel to achieve adequate signal to noise ratio. Recently, compressive sensing (CS) has been developed for single-pixel optical cameras to significantly reduce the imaging time and at the same time produce high-fidelity images by exploiting the sparsity of the data in some transform domain. While the efficacy of CS has been established for single-pixel optical systems, its application to PMMW imaging is not straightforward due to its (a) longer wavelength by three to four orders of magnitude that suffers high diffraction losses at finite size spatial waveform modulators and (b) weaker radiation intensity, for example, by eight orders of magnitude less than that of infrared. We present the development and implementation of a CS technique for PMMW imagers and shows a factor-of-ten increase in imaging speed. {\textcopyright} 2012 Society of Photo-Optical Instrumentation Engineers (SPIE).},
author = {Gopalsami, Nachappa and Liao, Shaolin and Elmer, Thomas W. and Koehl, Eugene R. and Heifetz, Alexander and Raptis, Apostolos C. and Spinoulas, Leonidas and Katsaggelos, Aggelos K.},
doi = {10.1117/1.OE.51.9.091614},
issn = {0091-3286},
journal = {Optical Engineering},
month = {sep},
number = {9},
pages = {091614--1},
title = {{Passive millimeter-wave imaging with compressive sensing}},
url = {http://opticalengineering.spiedigitallibrary.org/article.aspx?doi=10.1117/1.OE.51.9.091614},
volume = {51},
year = {2012}
}

Passive millimeter-wave (PMMW) imagers using a single radiometer, called single pixel imagers, employ raster scanning to produce images. A serious drawback of such a single pixel imaging system is the long acquisition time needed to produce a high-fidelity image, arising from two factors: (a) the time to scan the whole scene pixel by pixel and (b) the integration time for each pixel to achieve adequate signal to noise ratio. Recently, compressive sensing (CS) has been developed for single-pixel optical cameras to significantly reduce the imaging time and at the same time produce high-fidelity images by exploiting the sparsity of the data in some transform domain. While the efficacy of CS has been established for single-pixel optical systems, its application to PMMW imaging is not straightforward due to its (a) longer wavelength by three to four orders of magnitude that suffers high diffraction losses at finite size spatial waveform modulators and (b) weaker radiation intensity, for example, by eight orders of magnitude less than that of infrared. We present the development and implementation of a CS technique for PMMW imagers and shows a factor-of-ten increase in imaging speed. © 2012 Society of Photo-Optical Instrumentation Engineers (SPIE).

@article{Yun2012a,
author = {Fu, Yun and Hua, Xian-Sheng and Li, Zhu and Katsaggelos, Aggelos K. and Huang, Thomas S.},
doi = {10.1109/TSMCB.2011.2179292},
issn = {1083-4419},
journal = {IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics)},
month = {apr},
number = {2},
pages = {294--297},
pmid = {22262685},
title = {{Special Issue on Subspace and Manifold Learning for Image and Video Indexing and Search}},
url = {https://ieeexplore.ieee.org/document/6129536/},
volume = {42},
year = {2012}
}

@article{cortes2012variational,
abstract = {Although in the last decades the use of Magnetic Resonance Imaging has grown in popularity as a tool for the structural analysis of the brain, including MRI, fMRI and recently DTI, the ElectroEncephaloGraphy (EEG) is, still today, an interesting technique for the understanding of brain organization and function. The main reason for this is that the EEG is a direct measure of brain bioelectrical activity, and such activity can be monitorized in the millisecond time window. For some situations and cognitive scenarios, such fine temporal resolution might suffice for some aspects of brain function; however, the EEG spatial resolution is very poor since it is based on a small number of scalp recordings, thus turning the source localization problem into an ill-posed one in which infinite possibilities exist for the localization of the neuronal generators. This is an old problem in computational neuroimaging; indeed, many methods have been proposed to overcome this localization. Here, by performing a Variational Bayesian Inference procedure with a generalized Gaussian prior, we come out with an algorithm that performs simultaneously the estimation of both sources and model parameters. The novelty for the inclusion of the generalized Gaussian prior allows to control the smoothness degree of the estimated sources. Finally, the suggested algorithm is validated on simulated data.},
author = {Cortes, J. M. and Lopez, A. and Molina, R. and Katsaggelos, A. K.},
doi = {10.1140/epjp/i2012-12140-9},
issn = {2190-5444},
journal = {The European Physical Journal Plus},
month = {nov},
number = {11},
pages = {140},
publisher = {Springer Berlin Heidelberg},
title = {{Variational Bayesian localization of EEG sources with generalized Gaussian priors}},
url = {http://link.springer.com/10.1140/epjp/i2012-12140-9},
volume = {127},
year = {2012}
}

Although in the last decades the use of Magnetic Resonance Imaging has grown in popularity as a tool for the structural analysis of the brain, including MRI, fMRI and recently DTI, the ElectroEncephaloGraphy (EEG) is, still today, an interesting technique for the understanding of brain organization and function. The main reason for this is that the EEG is a direct measure of brain bioelectrical activity, and such activity can be monitorized in the millisecond time window. For some situations and cognitive scenarios, such fine temporal resolution might suffice for some aspects of brain function; however, the EEG spatial resolution is very poor since it is based on a small number of scalp recordings, thus turning the source localization problem into an ill-posed one in which infinite possibilities exist for the localization of the neuronal generators. This is an old problem in computational neuroimaging; indeed, many methods have been proposed to overcome this localization. Here, by performing a Variational Bayesian Inference procedure with a generalized Gaussian prior, we come out with an algorithm that performs simultaneously the estimation of both sources and model parameters. The novelty for the inclusion of the generalized Gaussian prior allows to control the smoothness degree of the estimated sources. Finally, the suggested algorithm is validated on simulated data.

@article{Lu2011,
abstract = {In this paper, we are concerned with image downsampling using subpixel techniques to achieve superior sharpness for small liquid crystal displays (LCDs). Such a problem exists when a high-resolution image or video is to be displayed on low-resolution display terminals. Limited by the low-resolution display, we have to shrink the image. Signal-processing theory tells us that optimal decimation requires low-pass filtering with a suitable cutoff frequency, followed by downsampling. In doing so, we need to remove many useful image details causing blurring. Subpixel-based downsampling, taking advantage of the fact that each pixel on a color LCD is actually composed of individual red, green, and blue subpixel stripes, can provide apparent higher resolution. In this paper, we use frequency-domain analysis to explain what happens in subpixel-based downsampling and why it is possible to achieve a higher apparent resolution. According to our frequency-domain analysis and observation, the cutoff frequency of the low-pass filter for subpixel-based decimation can be effectively extended beyond the Nyquist frequency using a novel antialiasing filter. Applying the proposed filters to two existing subpixel downsampling schemes called direct subpixel-based downsampling (DSD) and diagonal DSD (DDSD), we obtain two improved schemes, i.e., DSD based on frequency-domain analysis (DSD-FA) and DDSD based on frequency-domain analysis (DDSD-FA). Experimental results verify that the proposed DSD-FA and DDSD-FA can provide superior results, compared with existing subpixel or pixel-based downsampling methods. {\textcopyright} 2011 IEEE.},
author = {{Lu Fang} and Au, Oscar C. and {Ketan Tang} and Katsaggelos, Aggelos K. and Fang, Lu and Au, Oscar C. and Tang, Ketan and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2011.2165550},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Downsampling,frequency analysis,subpixel rendering},
month = {mar},
number = {3},
pages = {1391--1405},
title = {{Antialiasing Filter Design for Subpixel Downsampling via Frequency-Domain Analysis}},
url = {http://ieeexplore.ieee.org/document/5999717/},
volume = {21},
year = {2012}
}

In this paper, we are concerned with image downsampling using subpixel techniques to achieve superior sharpness for small liquid crystal displays (LCDs). Such a problem exists when a high-resolution image or video is to be displayed on low-resolution display terminals. Limited by the low-resolution display, we have to shrink the image. Signal-processing theory tells us that optimal decimation requires low-pass filtering with a suitable cutoff frequency, followed by downsampling. In doing so, we need to remove many useful image details causing blurring. Subpixel-based downsampling, taking advantage of the fact that each pixel on a color LCD is actually composed of individual red, green, and blue subpixel stripes, can provide apparent higher resolution. In this paper, we use frequency-domain analysis to explain what happens in subpixel-based downsampling and why it is possible to achieve a higher apparent resolution. According to our frequency-domain analysis and observation, the cutoff frequency of the low-pass filter for subpixel-based decimation can be effectively extended beyond the Nyquist frequency using a novel antialiasing filter. Applying the proposed filters to two existing subpixel downsampling schemes called direct subpixel-based downsampling (DSD) and diagonal DSD (DDSD), we obtain two improved schemes, i.e., DSD based on frequency-domain analysis (DSD-FA) and DDSD based on frequency-domain analysis (DDSD-FA). Experimental results verify that the proposed DSD-FA and DDSD-FA can provide superior results, compared with existing subpixel or pixel-based downsampling methods. © 2011 IEEE.

@article{Evaggelia2012a,
abstract = {In object-based video representation, video scenes are composed of several arbitrarily shaped video objects (VOs), defined by their texture, shape and motion. In error-prone communications, packet loss results in missing information at the decoder. The impact of transmission errors is minimized through error concealment. In this paper, we propose a spatial error concealment technique for recovering lost shape data. We consider a geometric shape representation consisting of the object boundary, which can be extracted from the $\alpha$-plane. Missing macroblocks result in a broken boundary. A B-spline curve is constructed to replace a missing boundary segment, based on a T-spline representation of the received boundary. We use T-splines because they produce shape-preserving approximations and do not change the characteristics of the original boundary. The representation ensures a good estimation of the first derivatives at the points touching the missing segment. Applying smoothing conditions, we manage to construct a new spline that joins smoothly with the received boundary, leading to successful concealment results. Experimental results on object shapes with different concealment difficulty demonstrate the performance of the proposed method. Comparisons with prior proposed methods are also presented. {\textcopyright} 1992-2012 IEEE.},
author = {Tsiligianni, Evaggelia and Kondi, Lisimachos P. and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2012.2192850},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {COM-ERC,Error concealment,Shape coding,T-splines},
month = {aug},
number = {8},
pages = {3573--3585},
pmid = {22481824},
title = {{Shape Error Concealment Based on a Shape-Preserving Boundary Approximation}},
url = {http://ieeexplore.ieee.org/document/6177257/},
volume = {21},
year = {2012}
}

In object-based video representation, video scenes are composed of several arbitrarily shaped video objects (VOs), defined by their texture, shape and motion. In error-prone communications, packet loss results in missing information at the decoder. The impact of transmission errors is minimized through error concealment. In this paper, we propose a spatial error concealment technique for recovering lost shape data. We consider a geometric shape representation consisting of the object boundary, which can be extracted from the $α$-plane. Missing macroblocks result in a broken boundary. A B-spline curve is constructed to replace a missing boundary segment, based on a T-spline representation of the received boundary. We use T-splines because they produce shape-preserving approximations and do not change the characteristics of the original boundary. The representation ensures a good estimation of the first derivatives at the points touching the missing segment. Applying smoothing conditions, we manage to construct a new spline that joins smoothly with the received boundary, leading to successful concealment results. Experimental results on object shapes with different concealment difficulty demonstrate the performance of the proposed method. Comparisons with prior proposed methods are also presented. © 1992-2012 IEEE.

@article{Naeem2012,
author = {Ramzan, Naeem and Izquierdo, Ebroul and Park, Hyunggon and Katsaggelos, Aggelos K. and Pouwelse, Johan},
doi = {10.1016/j.image.2012.04.003},
issn = {09235965},
journal = {Signal Processing: Image Communication},
month = {may},
number = {5},
pages = {379--382},
title = {{Special issue on advances in 2D/3D Video Streaming Over P2P Networks}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0923596512000823},
volume = {27},
year = {2012}
}

@article{Xiaoming2012,
abstract = {Purpose: To demonstrate the feasibility of using chemical shift magnetic resonance (MR) imaging fat-water separation methods for quantitative estimation of transcatheter lipiodol delivery to liver tissues. Materials and Methods: Studies were performed in accordance with institutional Animal Care and Use Committee guidelines. Proton nuclear MR spectroscopy was first performed to identify lipiodol spectral peaks and relative amplitudes. Next, phantoms were constructed with increasing lipiodol-water volume fractions. A multiecho chemical shift-based fatwater separation method was used to quantify lipiodol concentration within each phantom. Six rats served as controls; 18 rats underwent catheterization with digital subtraction angiography guidance for intraportal infusion of a 15%, 30%, or 50% by volume lipiodol-saline mixture. MR imaging measurements were used to quantify lipiodol delivery to each rat liver. Lipiodol concentration maps were reconstructed by using both single-peak and multipeak chemical shift models. Intraclass and Spearman correlation coefficients were calculated for statistical comparison of MR imaging-based lipiodol concentration and volume measurements to reference standards (known lipiodol phantom compositions and the infused lipiodol dose during rat studies). Results: Both single-peak and multipeak measurements were well correlated to phantom lipiodol concentrations (r 2 > 0.99). Lipiodol volume measurements were progressively and significantly higher when comparing between animals receiving different doses (P < .05 for each comparison). MR imaging-based lipiodol volume measurements strongly correlated with infused dose (intraclass correlation coefficients > 0.93, P < .001) with both single- and multipeak approaches. Conclusion: Chemical shift MR imaging fat-water separation methods can be used for quantitative measurements of lipiodol delivery to liver tissues. {\textcopyright} RSNA, 2012.},
author = {Yin, Xiaoming and Guo, Yang and Li, Weiguo and Huo, Eugene and Zhang, Zhuoli and Nicolai, Jodi and Kleps, Robert A. and Hernando, Diego and Katsaggelos, Aggelos K. and Omary, Reed A. and Larson, Andrew C.},
doi = {10.1148/radiol.12111916},
issn = {0033-8419},
journal = {Radiology},
month = {jun},
number = {3},
pages = {714--722},
pmid = {22623693},
title = {{Chemical Shift MR Imaging Methods for the Quantification of Transcatheter Lipiodol Delivery to the Liver: Preclinical Feasibility Studies in a Rodent Model}},
url = {http://pubs.rsna.org/doi/10.1148/radiol.12111916},
volume = {263},
year = {2012}
}

Purpose: To demonstrate the feasibility of using chemical shift magnetic resonance (MR) imaging fat-water separation methods for quantitative estimation of transcatheter lipiodol delivery to liver tissues. Materials and Methods: Studies were performed in accordance with institutional Animal Care and Use Committee guidelines. Proton nuclear MR spectroscopy was first performed to identify lipiodol spectral peaks and relative amplitudes. Next, phantoms were constructed with increasing lipiodol-water volume fractions. A multiecho chemical shift-based fatwater separation method was used to quantify lipiodol concentration within each phantom. Six rats served as controls; 18 rats underwent catheterization with digital subtraction angiography guidance for intraportal infusion of a 15%, 30%, or 50% by volume lipiodol-saline mixture. MR imaging measurements were used to quantify lipiodol delivery to each rat liver. Lipiodol concentration maps were reconstructed by using both single-peak and multipeak chemical shift models. Intraclass and Spearman correlation coefficients were calculated for statistical comparison of MR imaging-based lipiodol concentration and volume measurements to reference standards (known lipiodol phantom compositions and the infused lipiodol dose during rat studies). Results: Both single-peak and multipeak measurements were well correlated to phantom lipiodol concentrations (r 2 > 0.99). Lipiodol volume measurements were progressively and significantly higher when comparing between animals receiving different doses (P < .05 for each comparison). MR imaging-based lipiodol volume measurements strongly correlated with infused dose (intraclass correlation coefficients > 0.93, P < .001) with both single- and multipeak approaches. Conclusion: Chemical shift MR imaging fat-water separation methods can be used for quantitative measurements of lipiodol delivery to liver tissues. © RSNA, 2012.

@article{amizic2012sparse,
abstract = {In this article, we propose a novel blind image deconvolution method developed within the Bayesian framework. We concentrate on the restoration of blurred photographs taken by commercial cameras to show its effectiveness. The proposed method is based on a non-convex l p quasi norm with 0<p<1 that is used for the image, and a total variation (TV) based prior that is utilized for the blur. Bayesian inference is carried out by utilizing bounds for both the image and blur priors using a majorization-minimization principle. Maximum a posteriori estimates of the unknown image, blur and model parameters are calculated. Experimental results (i.e., restorations of more than 30 blurred photographs) are presented to demonstrate the advantage of the proposed method compared to existing ones. {\textcopyright} 2012 Amizic et al; licensee Springer.},
author = {Amizic, Bruno and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1186/1687-5281-2012-20},
issn = {16875176},
journal = {Eurasip Journal on Image and Video Processing},
number = {1},
pages = {1--15},
publisher = {SpringerOpen},
title = {{Sparse Bayesian blind image deconvolution with parameter estimation}},
volume = {2012},
year = {2012}
}

In this article, we propose a novel blind image deconvolution method developed within the Bayesian framework. We concentrate on the restoration of blurred photographs taken by commercial cameras to show its effectiveness. The proposed method is based on a non-convex l p quasi norm with 0

@article{babacan2012compressive,
abstract = {We propose a novel design for light field image acquisition based on compressive sensing principles. By placing a randomly coded mask at the aperture of a camera, incoherent measurements of the light passing through different parts of the lens are encoded in the captured images. Each captured image is a random linear combination of different angular views of a scene. The encoded images are then used to recover the original light field image via a novel Bayesian reconstruction algorithm. Using the principles of compressive sensing, we show that light field images with a large number of angular views can be recovered from only a few acquisitions. Moreover, the proposed acquisition and recovery method provides light field images with high spatial resolution and signal-to-noise-ratio, and therefore is not affected by limitations common to existing light field camera designs. We present a prototype camera design based on the proposed framework by modifying a regular digital camera. Finally, we demonstrate the effectiveness of the proposed system using experimental results with both synthetic and real images. {\textcopyright} 1992-2012 IEEE.},
author = {Babacan, S. Derin and Ansorge, Reto and Luessi, Martin and Mataran, Pablo Ruiz and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2012.2210237},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian methods,coded aperture,compressive sensing,computational photography,image reconstruction,light fields},
month = {dec},
number = {12},
pages = {4746--4757},
publisher = {IEEE},
title = {{Compressive Light Field Sensing}},
url = {http://ieeexplore.ieee.org/document/6248701/},
volume = {21},
year = {2012}
}

@inproceedings{lopez2012hyperparameters,
author = {L{\'{o}}pez, Antonio and Cort{\'{e}}s, Jes{\'{u}}s M and L{\'{o}}pez-Oiler, Domingo and Molina, Rafael and Katsaggelos, Aggelos K},
booktitle = {2012 Proceedings of the 20th European Signal Processing Conference (EUSIPCO)},
organization = {IEEE},
pages = {489--493},
title = {{Hyperparameters estimation for the Bayesian localization of the EEG sources with TV priors}},
year = {2012}
}

@article{babacan2012compressive,
abstract = {We propose a novel design for light field image acquisition based on compressive sensing principles. By placing a randomly coded mask at the aperture of a camera, incoherent measurements of the light passing through different parts of the lens are encoded in the captured images. Each captured image is a random linear combination of different angular views of a scene. The encoded images are then used to recover the original light field image via a novel Bayesian reconstruction algorithm. Using the principles of compressive sensing, we show that light field images with a large number of angular views can be recovered from only a few acquisitions. Moreover, the proposed acquisition and recovery method provides light field images with high spatial resolution and signal-to-noise-ratio, and therefore is not affected by limitations common to existing light field camera designs. We present a prototype camera design based on the proposed framework by modifying a regular digital camera. Finally, we demonstrate the effectiveness of the proposed system using experimental results with both synthetic and real images. {\textcopyright} 1992-2012 IEEE.},
author = {Babacan, S. Derin and Ansorge, Reto and Luessi, Martin and Mataran, Pablo Ruiz and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2012.2210237},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian methods,coded aperture,compressive sensing,computational photography,image reconstruction,light fields},
month = {dec},
number = {12},
pages = {4746--4757},
publisher = {IEEE},
title = {{Compressive Light Field Sensing}},
url = {http://ieeexplore.ieee.org/document/6248701/},
volume = {21},
year = {2012}
}

We propose a novel design for light field image acquisition based on compressive sensing principles. By placing a randomly coded mask at the aperture of a camera, incoherent measurements of the light passing through different parts of the lens are encoded in the captured images. Each captured image is a random linear combination of different angular views of a scene. The encoded images are then used to recover the original light field image via a novel Bayesian reconstruction algorithm. Using the principles of compressive sensing, we show that light field images with a large number of angular views can be recovered from only a few acquisitions. Moreover, the proposed acquisition and recovery method provides light field images with high spatial resolution and signal-to-noise-ratio, and therefore is not affected by limitations common to existing light field camera designs. We present a prototype camera design based on the proposed framework by modifying a regular digital camera. Finally, we demonstrate the effectiveness of the proposed system using experimental results with both synthetic and real images. © 1992-2012 IEEE.

@article{Miguel2011,
abstract = {In this paper we propose a novel super-resolution based algorithm for the pansharpening of multispectral images. Within the Bayesian formulation, the proposed methodology incorporates prior knowledge on the expected characteristics of multispectral images; that is, it imposes smoothness within each band by means of the energy associated with the ℓ1 norm of vertical and horizontal first order differences of image pixel values and also takes into account the correlation among the bands of the multispectral image. The observation process is modeled using the sensor characteristics of both panchromatic and multispectral images. The method is tested on real and synthetic images, compared with other pansharpening methods, and the quality of the results assessed both qualitatively and quantitatively. {\textcopyright} 2010 Springer Science+Business Media, LLC.},
author = {Vega, Miguel and Mateos, Javier and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1007/s11265-010-0554-x},
issn = {1939-8018},
journal = {Journal of Signal Processing Systems},
keywords = {Bayesian approach,Interband correlations,Multispectral images,Pansharpening,Super-resolution,Variational methods,ℓ1 image models},
month = {dec},
number = {3},
pages = {509--523},
title = {{Super Resolution of Multispectral Images using ℓ1 Image Models and Interband Correlations}},
url = {http://link.springer.com/10.1007/s11265-010-0554-x},
volume = {65},
year = {2011}
}

In this paper we propose a novel super-resolution based algorithm for the pansharpening of multispectral images. Within the Bayesian formulation, the proposed methodology incorporates prior knowledge on the expected characteristics of multispectral images; that is, it imposes smoothness within each band by means of the energy associated with the ℓ1 norm of vertical and horizontal first order differences of image pixel values and also takes into account the correlation among the bands of the multispectral image. The observation process is modeled using the sensor characteristics of both panchromatic and multispectral images. The method is tested on real and synthetic images, compared with other pansharpening methods, and the quality of the results assessed both qualitatively and quantitatively. © 2010 Springer Science+Business Media, LLC.

@article{Eren2011,
abstract = {In centralized transportation surveillance systems, video is captured and compressed at low processing power remote nodes and transmitted to a central location for processing. Such compression can reduce the accuracy of centrally run automated object tracking algorithms. In typical systems, the majority of communications bandwidth is spent on encoding temporal pixel variations such as acquisition noise or local changes to lighting. We propose a tracking-aware, H.264-compliant compression algorithm that removes temporal components of low tracking interest and optimizes the quantization of frequency coefficients, particularly those that most influence trackers, significantly reducing bitrate while maintaining comparable tracking accuracy. We utilize tracking accuracy as our compression criterion in lieu of mean squared error metrics. Our proposed system is designed with low processing power and memory requirements in mind, and as such can be deployed on remote nodes. Using H.264/AVC video coding and a commonly used state-of-the-art tracker we show that our algorithm allows for over 90% bitrate savings while maintaining comparable tracking accuracy. {\textcopyright} 2011 IEEE.},
author = {Soyak, Eren and Tsaftaris, Sotirios A. and Katsaggelos, Aggelos K.},
doi = {10.1109/TCSVT.2011.2163448},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
keywords = {Quantization,urban transportation video,video compression,video object tracking,video processing},
month = {oct},
number = {10},
pages = {1378--1389},
title = {{Low-Complexity Tracking-Aware H.264 Video Compression for Transportation Surveillance}},
url = {http://ieeexplore.ieee.org/document/5971775/},
volume = {21},
year = {2011}
}

In centralized transportation surveillance systems, video is captured and compressed at low processing power remote nodes and transmitted to a central location for processing. Such compression can reduce the accuracy of centrally run automated object tracking algorithms. In typical systems, the majority of communications bandwidth is spent on encoding temporal pixel variations such as acquisition noise or local changes to lighting. We propose a tracking-aware, H.264-compliant compression algorithm that removes temporal components of low tracking interest and optimizes the quantization of frequency coefficients, particularly those that most influence trackers, significantly reducing bitrate while maintaining comparable tracking accuracy. We utilize tracking accuracy as our compression criterion in lieu of mean squared error metrics. Our proposed system is designed with low processing power and memory requirements in mind, and as such can be deployed on remote nodes. Using H.264/AVC video coding and a commonly used state-of-the-art tracker we show that our algorithm allows for over 90% bitrate savings while maintaining comparable tracking accuracy. © 2011 IEEE.

@article{babacan2017variational,
abstract = {In this paper, we address the super resolution (SR) problem from a set of degraded low resolution (LR) images to obtain a high resolution (HR) image. Accurate estimation of the sub-pixel motion between the LR images significantly affects the performance of the reconstructed HR image. In this paper, we propose novel super resolution methods where the HR image and the motion parameters are estimated simultaneously. Utilizing a Bayesian formulation, we model the unknown HR image, the acquisition process, the motion parameters and the unknown model parameters in a stochastic sense. Employing a variational Bayesian analysis, we develop two novel algorithms which jointly estimate the distributions of all unknowns. The proposed framework has the following advantages: 1) Through the incorporation of uncertainty of the estimates, the algorithms prevent the propagation of errors between the estimates of the various unknowns; 2) the algorithms are robust to errors in the estimation of the motion parameters; and 3) using a fully Bayesian formulation, the developed algorithms simultaneously estimate all algorithmic parameters along with the HR image and motion parameters, and therefore they are fully-automated and do not require parameter tuning. We also show that the proposed motion estimation method is a stochastic generalization of the classical Lucas-Kanade registration algorithm. Experimental results demonstrate that the proposed approaches are very effective and compare favorably to state-of-the-art SR algorithms. {\textcopyright} 2011 IEEE.},
author = {Babacan, S. Derin and Molina, Rafael and Katsaggelos, Aggelos K. and {Derin Babacan}, S. and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2010.2080278},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian methods,parameter estimation,super resolution,total variation,variational methods},
month = {apr},
number = {4},
pages = {984--999},
pmid = {20876021},
publisher = {CRC Press},
title = {{Variational Bayesian Super Resolution}},
url = {https://www.taylorfrancis.com/chapters/edit/10.1201/9781439819319-10/variational-bayesian-super-resolution-reconstruction-derin-babacan-rafael-molina-aggelos-katsaggelos https://www.taylorfrancis.com/books/9781439819319/chapters/10.1201/9781439819319-10 h},
volume = {20},
year = {2011}
}

In this paper, we address the super resolution (SR) problem from a set of degraded low resolution (LR) images to obtain a high resolution (HR) image. Accurate estimation of the sub-pixel motion between the LR images significantly affects the performance of the reconstructed HR image. In this paper, we propose novel super resolution methods where the HR image and the motion parameters are estimated simultaneously. Utilizing a Bayesian formulation, we model the unknown HR image, the acquisition process, the motion parameters and the unknown model parameters in a stochastic sense. Employing a variational Bayesian analysis, we develop two novel algorithms which jointly estimate the distributions of all unknowns. The proposed framework has the following advantages: 1) Through the incorporation of uncertainty of the estimates, the algorithms prevent the propagation of errors between the estimates of the various unknowns; 2) the algorithms are robust to errors in the estimation of the motion parameters; and 3) using a fully Bayesian formulation, the developed algorithms simultaneously estimate all algorithmic parameters along with the HR image and motion parameters, and therefore they are fully-automated and do not require parameter tuning. We also show that the proposed motion estimation method is a stochastic generalization of the classical Lucas-Kanade registration algorithm. Experimental results demonstrate that the proposed approaches are very effective and compare favorably to state-of-the-art SR algorithms. © 2011 IEEE.

@article{amro2011survey,
abstract = {There exist a number of satellites on different earth observation platforms, which provide multispectral images together with a panchromatic image, that is, an image containing reflectance data representative of a wide range of bands and wavelengths. Pansharpening is a pixel-level fusion technique used to increase the spatial resolution of the multispectral image while simultaneously preserving its spectral information. In this paper, we provide a review of the pan-sharpening methods proposed in the literature giving a clear classification of them and a description of their main characteristics. Finally, we analyze how the quality of the pansharpened images can be assessed both visually and quantitatively and examine the different quality measures proposed for that purpose.},
author = {Amro, Israa and Mateos, Javier and Vega, Miguel and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1186/1687-6180-2011-79},
issn = {1687-6180},
journal = {EURASIP Journal on Advances in Signal Processing},
month = {dec},
number = {1},
pages = {79},
publisher = {SpringerOpen},
title = {{A survey of classical methods and new trends in pansharpening of multispectral images}},
url = {https://asp-eurasipjournals.springeropen.com/articles/10.1186/1687-6180-2011-79},
volume = {2011},
year = {2011}
}

There exist a number of satellites on different earth observation platforms, which provide multispectral images together with a panchromatic image, that is, an image containing reflectance data representative of a wide range of bands and wavelengths. Pansharpening is a pixel-level fusion technique used to increase the spatial resolution of the multispectral image while simultaneously preserving its spectral information. In this paper, we provide a review of the pan-sharpening methods proposed in the literature giving a clear classification of them and a description of their main characteristics. Finally, we analyze how the quality of the pansharpened images can be assessed both visually and quantitatively and examine the different quality measures proposed for that purpose.

@article{Sotirios2011,
author = {Tsaftaris, Sotirios and Lister, Kristin and Fiedler, Inge and Casadio, Francesca and Katsaggelos, Aggelos},
doi = {10.1109/MSP.2011.940408},
issn = {1053-5888},
journal = {IEEE Signal Processing Magazine},
month = {may},
number = {3},
pages = {113--119},
title = {{Colorizing a Masterpiece [Applications Corner]}},
url = {http://ieeexplore.ieee.org/document/5753071/},
volume = {28},
year = {2011}
}

@article{Martin2011,
abstract = {In this paper, we propose a novel symmetrical EEG/fMRI fusion method which combines EEG and fMRI by means of a common generative model. We use a total variation (TV) prior to model the spatial distribution of the cortical current responses and hemodynamic response functions, and utilize spatially adaptive temporal priors to model their temporal shapes. The spatial adaptivity of the prior model allows for adaptation to the local characteristics of the estimated responses and leads to high estimation performance for the cortical current distribution and the hemodynamic response functions. We utilize a Bayesian formulation with a variational Bayesian framework and obtain a fully automatic fusion algorithm. Simulations with synthetic data and experiments with real data from a multimodal study on face perception demonstrate the performance of the proposed method. {\textcopyright} 2010 Elsevier Inc.},
author = {Luessi, Martin and Babacan, S. Derin and Molina, Rafael and Booth, James R. and Katsaggelos, Aggelos K.},
doi = {10.1016/j.neuroimage.2010.11.037},
issn = {10538119},
journal = {NeuroImage},
keywords = {M/EEG source localization,Multimodal fusion,Spatial adaptivity,Total variation,Variational Bayes},
month = {mar},
number = {1},
pages = {113--132},
pmid = {21130173},
title = {{Bayesian symmetrical EEG/fMRI fusion with spatially adaptive priors}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1053811910014679},
volume = {55},
year = {2011}
}

In this paper, we propose a novel symmetrical EEG/fMRI fusion method which combines EEG and fMRI by means of a common generative model. We use a total variation (TV) prior to model the spatial distribution of the cortical current responses and hemodynamic response functions, and utilize spatially adaptive temporal priors to model their temporal shapes. The spatial adaptivity of the prior model allows for adaptation to the local characteristics of the estimated responses and leads to high estimation performance for the cortical current distribution and the hemodynamic response functions. We utilize a Bayesian formulation with a variational Bayesian framework and obtain a fully automatic fusion algorithm. Simulations with synthetic data and experiments with real data from a multimodal study on face perception demonstrate the performance of the proposed method. © 2010 Elsevier Inc.

@article{Fan2011,
abstract = {Compared to other anomalous video event detection approaches that analyze object trajectories only, we propose a context-aware method to detect anomalies. By tracking all moving objects in the video, three different levels of spatiotemporal contexts are considered, i.e., point anomaly of a video object, sequential anomaly of an object trajectory, and co-occurrence anomaly of multiple video objects. A hierarchical data mining approach is proposed. At each level, frequency-based analysis is performed to automatically discover regular rules of normal events. Events deviating from these rules are identified as anomalies. The proposed method is computationally efficient and can infer complex rules. Experiments on real traffic video validate that the detected video anomalies are hazardous or illegal according to traffic regulations. {\textcopyright} 2010 Elsevier Inc. All rights reserved.},
author = {Jiang, Fan and Yuan, Junsong and Tsaftaris, Sotirios A. and Katsaggelos, Aggelos K.},
doi = {10.1016/j.cviu.2010.10.008},
issn = {10773142},
journal = {Computer Vision and Image Understanding},
keywords = {Anomaly detection,Clustering,Context,Data mining,Video surveillance},
month = {mar},
number = {3},
pages = {323--333},
title = {{Anomalous video event detection using spatiotemporal context}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1077314210002390},
volume = {115},
year = {2011}
}

Compared to other anomalous video event detection approaches that analyze object trajectories only, we propose a context-aware method to detect anomalies. By tracking all moving objects in the video, three different levels of spatiotemporal contexts are considered, i.e., point anomaly of a video object, sequential anomaly of an object trajectory, and co-occurrence anomaly of multiple video objects. A hierarchical data mining approach is proposed. At each level, frequency-based analysis is performed to automatically discover regular rules of normal events. Events deviating from these rules are identified as anomalies. The proposed method is computationally efficient and can infer complex rules. Experiments on real traffic video validate that the detected video anomalies are hazardous or illegal according to traffic regulations. © 2010 Elsevier Inc. All rights reserved.

@article{babacan2010bayesian,
abstract = {Photographs acquired under low-lighting conditions require long exposure times and therefore exhibit significant blurring due to the shaking of the camera. Using shorter exposure times results in sharper images but with a very high level of noise. In this paper, we address the problem of utilizing two such images in order to obtain an estimate of the original scene and present a novel blind deconvolution algorithm for solving it. We formulate the problem in a hierarchical Bayesian framework by utilizing prior knowledge on the unknown image and blur, and also on the dependency between the two observed images. By incorporating a fully Bayesian analysis, the developed algorithm estimates all necessary model parameters along with the unknown image and blur, such that no user-intervention is needed. Moreover, we employ a variational Bayesian inference procedure, which allows for the statistical compensation of errors occurring at different stages of the restoration, and also provides uncertainties of the estimates. Experimental results with synthetic and real images demonstrate that the proposed method provides very high quality restoration results and compares favorably to existing methods even though no user supervision is needed. {\textcopyright} 2006 IEEE.},
author = {Babacan, Sevket Derin and {Jingnan Wang} and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2010.2052263},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian methods,blind deconvolution,image stabilization,parameter estimation,variational distribution approximations},
month = {nov},
number = {11},
pages = {2874--2888},
publisher = {IEEE},
title = {{Bayesian Blind Deconvolution From Differently Exposed Image Pairs}},
url = {http://ieeexplore.ieee.org/document/5482163/},
volume = {19},
year = {2010}
}

Photographs acquired under low-lighting conditions require long exposure times and therefore exhibit significant blurring due to the shaking of the camera. Using shorter exposure times results in sharper images but with a very high level of noise. In this paper, we address the problem of utilizing two such images in order to obtain an estimate of the original scene and present a novel blind deconvolution algorithm for solving it. We formulate the problem in a hierarchical Bayesian framework by utilizing prior knowledge on the unknown image and blur, and also on the dependency between the two observed images. By incorporating a fully Bayesian analysis, the developed algorithm estimates all necessary model parameters along with the unknown image and blur, such that no user-intervention is needed. Moreover, we employ a variational Bayesian inference procedure, which allows for the statistical compensation of errors occurring at different stages of the restoration, and also provides uncertainties of the estimates. Experimental results with synthetic and real images demonstrate that the proposed method provides very high quality restoration results and compares favorably to existing methods even though no user supervision is needed. © 2006 IEEE.

@article{Sotirios2010a,
author = {Tsaftaris, Sotirios A and Zhou, Xiangzhi and Tang, Richard and Katsaggelos, Aggelos and Li, Debiao and Dharmakumar, Rohan},
doi = {10.1186/1532-429X-12-S1-P216},
issn = {1532-429X},
journal = {Journal of Cardiovascular Magnetic Resonance},
month = {jan},
number = {S1},
pages = {P216},
title = {{Automated detection and quantification of microcirculatory oxygenation changes in the heart}},
url = {https://jcmr-online.biomedcentral.com/articles/10.1186/1532-429X-12-S1-P216},
volume = {12},
year = {2010}
}

@article{Dalei2010,
abstract = {Most existing works on routing for video transmission over multihop wireless networks only focus on how to satisfy the network-oriented quality-of-service (QoS), such as throughput, delay, and packet loss rate rather than application-oriented QoS such as the user-perceived video quality. Although there are some research efforts which use application-centric video quality as the routing metric, they either calculate the video quality based on some predefined rate-distortion function or model without considering the impact of video coding and decoding (including error concealment) on routing, or use exhaustive search or heuristic methods to find the optimal path, leading to high computational complexity and/or suboptimal solutions. In this paper, we propose an application-centric routing framework for real-time video transmission in multihop wireless networks, where expected video distortion is adopted as the routing metric. The major contributions of this paper are: 1) the development of an efficient routing algorithm with the routing metric expressed in terms of the expected video distortion and being calculated on-the-fly, and 2) the development of a quality-driven cross-layer optimization framework to enhance the flexibility and robustness of routing by the joint optimization of routing path selection and video coding, thereby maximizing the user-perceived video quality under a given video playback delay constraint. Both theoretical and experimental results demonstrate that the proposed quality-driven application-centric routing approach can achieve a superior performance over existing network-centric routing approaches. {\textcopyright} 2006 IEEE.},
author = {Wu, Dalei and Ci, Song and Wang, Haohong and Katsaggelos, Aggelos K.},
doi = {10.1109/TCSVT.2010.2057014},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
keywords = {Cross-layer optimization,dynamic programming,error concealment,multihop wireless networks,routing,video coding},
month = {dec},
number = {12},
pages = {1721--1734},
title = {{Application-Centric Routing for Video Streaming Over MultiHop Wireless Networks}},
url = {http://ieeexplore.ieee.org/document/5508380/},
volume = {20},
year = {2010}
}

Most existing works on routing for video transmission over multihop wireless networks only focus on how to satisfy the network-oriented quality-of-service (QoS), such as throughput, delay, and packet loss rate rather than application-oriented QoS such as the user-perceived video quality. Although there are some research efforts which use application-centric video quality as the routing metric, they either calculate the video quality based on some predefined rate-distortion function or model without considering the impact of video coding and decoding (including error concealment) on routing, or use exhaustive search or heuristic methods to find the optimal path, leading to high computational complexity and/or suboptimal solutions. In this paper, we propose an application-centric routing framework for real-time video transmission in multihop wireless networks, where expected video distortion is adopted as the routing metric. The major contributions of this paper are: 1) the development of an efficient routing algorithm with the routing metric expressed in terms of the expected video distortion and being calculated on-the-fly, and 2) the development of a quality-driven cross-layer optimization framework to enhance the flexibility and robustness of routing by the joint optimization of routing path selection and video coding, thereby maximizing the user-perceived video quality under a given video playback delay constraint. Both theoretical and experimental results demonstrate that the proposed quality-driven application-centric routing approach can achieve a superior performance over existing network-centric routing approaches. © 2006 IEEE.

@article{T2010,
author = {T, Adali and A, Bovik and F, Juang and A, Katsaggelos and A, C Kot and J, Krolik and Kjray, L I U and D, Malah and V, J Mathews and H, Messer-Yaron and Et al.},
doi = {10.1109/JSTSP.2010.2040787},
issn = {1932-4553},
journal = {IEEE Journal of Selected Topics in Signal Processing},
month = {feb},
number = {1},
pages = {C2--C2},
title = {{IEEE Journal of Selected Topics in Signal Processing publication information}},
url = {http://ieeexplore.ieee.org/document/5393283/},
volume = {4},
year = {2010}
}

@article{Xiaoming2010,
abstract = {Accurate R2* measurements are critical for many abdominal imaging applications. Conventionally, R2* maps are derived via the monoexponential fitting of signal decay within a series of gradient-echo (GRE) images reconstructed from multichannel datasets combined using a root sum-of-squares (RSS) approach. However, the noise bias at low-SNR TEs from RSS-reconstructed data often causes the underestimation of R2* values. In phantom, ex vivo animal model and normal volunteer studies, we investigated the accuracy of low-SNR R2* measurement when combining truncation and coil combination methods. The accuracy for R2* estimations was shown to be affected by the intrinsic R2* value, SNR level and the chosen reconstruction method. The R2* estimation error was found to decrease with increasing SNR level, decreasing R2* value and the use of the optimal B1-weighted combined (OBC) image reconstruction method. Data truncation based on rigorous voxel-wise SNR estimates can reduce R2* measurement error in the setting of low SNR with fast signal decay. When optimal SNR truncation thresholds are unknown, the OBC method can provide optimal R2* measurements given the minimal truncation requirements. {\textcopyright} 2010 John Wiley & Sons, Ltd.},
author = {Yin, Xiaoming and Shah, Saurabh and Katsaggelos, Aggelos K. and Larson, Andrew C.},
doi = {10.1002/nbm.1539},
issn = {09523480},
journal = {NMR in Biomedicine},
keywords = {Multiple gradient-echo,Noise bias,Optimal B1-weighted image reconstruction,Phase array coils,R2* mapping,Root sum-of-square,Signal to noise ratios,Truncation},
month = {dec},
number = {10},
pages = {1127--1136},
pmid = {21162142},
title = {{Improved R2* measurement accuracy with absolute SNR truncation and optimal coil combination}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/nbm.1539},
volume = {23},
year = {2010}
}

Accurate R2* measurements are critical for many abdominal imaging applications. Conventionally, R2* maps are derived via the monoexponential fitting of signal decay within a series of gradient-echo (GRE) images reconstructed from multichannel datasets combined using a root sum-of-squares (RSS) approach. However, the noise bias at low-SNR TEs from RSS-reconstructed data often causes the underestimation of R2* values. In phantom, ex vivo animal model and normal volunteer studies, we investigated the accuracy of low-SNR R2* measurement when combining truncation and coil combination methods. The accuracy for R2* estimations was shown to be affected by the intrinsic R2* value, SNR level and the chosen reconstruction method. The R2* estimation error was found to decrease with increasing SNR level, decreasing R2* value and the use of the optimal B1-weighted combined (OBC) image reconstruction method. Data truncation based on rigorous voxel-wise SNR estimates can reduce R2* measurement error in the setting of low SNR with fast signal decay. When optimal SNR truncation thresholds are unknown, the OBC method can provide optimal R2* measurements given the minimal truncation requirements. © 2010 John Wiley & Sons, Ltd.

@article{chantas2009variational,
abstract = {In this paper, a new image prior is introduced and used in image restoration. This prior is based on products of spatially weighted total variations (TV). These spatial weights provide this prior with the flexibility to better capture local image features than previous TV based priors. Bayesian inference is used for image restoration with this prior via the variational approximation. The proposed restoration algorithm is fully automatic in the sense that all necessary parameters are estimated from the data and is faster than previous similar algorithms. Numerical experiments are shown which demonstrate that image restoration based on this prior compares favorably with previous state-of-the-art restoration algorithms. {\textcopyright} 2010 IEEE.},
author = {Chantas, Giannis and Galatsanos, Nikolaos P. and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2009.2033398},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {No Keywords.},
month = {feb},
number = {2},
pages = {351--362},
publisher = {IEEE},
title = {{Variational Bayesian Image Restoration With a Product of Spatially Weighted Total Variation Image Priors}},
url = {http://ieeexplore.ieee.org/document/5272318/},
volume = {19},
year = {2010}
}

In this paper, a new image prior is introduced and used in image restoration. This prior is based on products of spatially weighted total variations (TV). These spatial weights provide this prior with the flexibility to better capture local image features than previous TV based priors. Bayesian inference is used for image restoration with this prior via the variational approximation. The proposed restoration algorithm is fully automatic in the sense that all necessary parameters are estimated from the data and is faster than previous similar algorithms. Numerical experiments are shown which demonstrate that image restoration based on this prior compares favorably with previous state-of-the-art restoration algorithms. © 2010 IEEE.

@article{Jiucai2009,
author = {Jiucai, Zhang and Dalei, Wu and Song, Ci and Haohong, Wang and Aggelos, K Katsaggelos},
journal = {J. Commun.},
pages = {600--613},
title = {{Power-aware mobile multimedia: a survey}},
volume = {4},
year = {2009}
}

@article{LauraAnn2009,
abstract = {Hearing aid users have difficulty hearing target signals, such as speech, in the presence of competing signals or noise. Most solutions proposed to date enhance or extract target signals from background noise and interference based on either location attributes or source attributes. Location attributes typically involve arrival angles at a microphone array. Source attributes include characteristics that are specific to a signal, such as fundamental frequency, or statistical properties that differentiate signals. This paper describes a novel approach to sound source separation, called computational auditory scene analysis-enhanced beamforming (CASA-EB), that achieves increased separation performance by combining the complementary techniques of CASA (a source attribute technique) with beamforming (a location attribute technique), complementary in the sense that they use independent attributes for signal separation. CASA-EB performs sound source separation by temporally and spatially filtering a multichannel input signal, and then grouping the resulting signal components into separated signals, based on source and location attributes. Experimental results show increased signal-to-interference ratio with CASA-EB over beamforming or CASA alone. Copyright {\textcopyright} 2009 L. A. Drake et al.},
author = {Drake, L. A. and Rutledge, J. C. and Zhang, J. and Katsaggelos, A.},
doi = {10.1155/2009/403681},
issn = {1687-6180},
journal = {EURASIP Journal on Advances in Signal Processing},
month = {dec},
number = {1},
pages = {403681},
title = {{A Computational Auditory Scene Analysis-Enhanced Beamforming Approach for Sound Source Separation}},
url = {https://asp-eurasipjournals.springeropen.com/articles/10.1155/2009/403681},
volume = {2009},
year = {2009}
}

Hearing aid users have difficulty hearing target signals, such as speech, in the presence of competing signals or noise. Most solutions proposed to date enhance or extract target signals from background noise and interference based on either location attributes or source attributes. Location attributes typically involve arrival angles at a microphone array. Source attributes include characteristics that are specific to a signal, such as fundamental frequency, or statistical properties that differentiate signals. This paper describes a novel approach to sound source separation, called computational auditory scene analysis-enhanced beamforming (CASA-EB), that achieves increased separation performance by combining the complementary techniques of CASA (a source attribute technique) with beamforming (a location attribute technique), complementary in the sense that they use independent attributes for signal separation. CASA-EB performs sound source separation by temporally and spatially filtering a multichannel input signal, and then grouping the resulting signal components into separated signals, based on source and location attributes. Experimental results show increased signal-to-interference ratio with CASA-EB over beamforming or CASA alone. Copyright © 2009 L. A. Drake et al.

@article{babacan2008variational,
abstract = {In this paper, we present novel algorithms for total variation (TV) based blind deconvolution and parameter estimation utilizing a variational framework. Using a hierarchical Bayesian model, the unknown image, blur, and hyperparameters for the image, blur, and noise priors are estimated simultaneously. A variational inference approach is utilized so that approximations of the posterior distributions of the unknowns are obtained, thus providing a measure of the uncertainty of the estimates. Experimental results demonstrate that the proposed approaches provide higher restoration performance than non-TV-based methods without any assumptions about the unknown hyperparameters. {\textcopyright} 2008 IEEE.},
author = {Babacan, S. Derin and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2008.2007354},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian methods,Blind deconvolution,Parameter estimation,Total variation (TV),Variational methods},
number = {1},
pages = {12--26},
pmid = {19095515},
publisher = {IEEE},
title = {{Variational Bayesian blind deconvolution using a total variation prior}},
volume = {18},
year = {2009}
}

In this paper, we present novel algorithms for total variation (TV) based blind deconvolution and parameter estimation utilizing a variational framework. Using a hierarchical Bayesian model, the unknown image, blur, and hyperparameters for the image, blur, and noise priors are estimated simultaneously. A variational inference approach is utilized so that approximations of the posterior distributions of the unknowns are obtained, thus providing a measure of the uncertainty of the estimates. Experimental results demonstrate that the proposed approaches provide higher restoration performance than non-TV-based methods without any assumptions about the unknown hyperparameters. © 2008 IEEE.

@article{Konstantinos2009,
abstract = {A new, reduced complexity algorithm is proposed for compensating the Inter-Carrier Interference (ICI) caused by severe PHase Noise (PHN) and Residual Frequency Offset (RFO) in OFDM systems. The algorithm estimates and compensates the most significant terms of the frequency domain ICI process, which are optimally selected via a Minimum Mean Squared Error (MMSE) criterion. The algorithm requires minimal knowledge of the phase process statistics, the estimation of which is also considered. The scheme outperforms previously proposed compensation methods of similar complexity, when severe phase impairments are present. {\textcopyright} 2006 IEEE.},
author = {Nikitopoulos, Konstantinos and Stefanatos, Stelios and Katsaggelos, A.K.},
doi = {10.1109/TWC.2009.071029},
issn = {1536-1276},
journal = {IEEE Transactions on Wireless Communications},
keywords = {Frequency offset,Inter-carrier interference,OFDM,Phase noise},
month = {apr},
number = {4},
pages = {1614--1619},
title = {{Decision-aided compensation of severe phase-impairment-induced inter-carrier interference in frequency-selective OFDM}},
url = {http://ieeexplore.ieee.org/document/4907430/},
volume = {8},
year = {2009}
}

A new, reduced complexity algorithm is proposed for compensating the Inter-Carrier Interference (ICI) caused by severe PHase Noise (PHN) and Residual Frequency Offset (RFO) in OFDM systems. The algorithm estimates and compensates the most significant terms of the frequency domain ICI process, which are optimally selected via a Minimum Mean Squared Error (MMSE) criterion. The algorithm requires minimal knowledge of the phase process statistics, the estimation of which is also considered. The scheme outperforms previously proposed compensation methods of similar complexity, when severe phase impairments are present. © 2006 IEEE.

@article{jiang2009dynamic,
abstract = {The proposed unusual video event detection method is based on unsupervised clustering of object trajectories, which are modeled by hidden Markov models (HMM). The novelty of the method includes a dynamic hierarchical process incorporated in the trajectory clustering algorithm to prevent model overfitting and a 2-depth greedy search strategy for efficient clustering. {\textcopyright} 2009 IEEE.},
author = {{Fan Jiang} and {Ying Wu} and Katsaggelos, Aggelos K A.K. and Jiang, Fan and Wu, Ying and Katsaggelos, Aggelos K A.K.},
doi = {10.1109/TIP.2008.2012070},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Event detection,Unsupervised clustering,Video surveillance},
month = {apr},
number = {4},
pages = {907--913},
publisher = {IEEE},
title = {{A dynamic hierarchical clustering method for trajectory-based unusual video event detection}},
url = {http://ieeexplore.ieee.org/document/4798178/},
volume = {18},
year = {2009}
}

The proposed unusual video event detection method is based on unsupervised clustering of object trajectories, which are modeled by hidden Markov models (HMM). The novelty of the method includes a dynamic hierarchical process incorporated in the trajectory clustering algorithm to prevent model overfitting and a 2-depth greedy search strategy for efficient clustering. © 2009 IEEE.

@article{gao2009efficient,
abstract = {With the phenomenal growth of the online and personal video repositories, an efficient and robust example-based video search solution is required to support applications like query by clip, query by capture, and repeated clip detection. In this letter, video sequences are represented as temporal trajectories via scaling and lower dimensional representation of the video frame luminance field, and a video trajectory indexing and matching scheme is developed to support video clip search. Simulation results demonstrate that the proposed approach achieves excellent performance in both response speed and precision-recall accuracy. {\textcopyright} 2009 IEEE.},
author = {Gao, Li and Li, Zhu and Katsaggelos, Aggelos and {Li Gao} and {Zhu Li} and Katsaggelos, Aggelos},
doi = {10.1109/TCSVT.2009.2026813},
issn = {1051-8215},
journal = {IEEE transactions on circuits and systems for video technology},
keywords = {Modeling,Subspace learning,Video indexing,Video retrieval},
month = {oct},
number = {10},
pages = {1566--1570},
publisher = {IEEE},
title = {{An Efficient Video Indexing and Retrieval Algorithm Using the Luminance Field Trajectory Modeling}},
url = {http://ieeexplore.ieee.org/document/5159407/},
volume = {19},
year = {2009}
}

With the phenomenal growth of the online and personal video repositories, an efficient and robust example-based video search solution is required to support applications like query by clip, query by capture, and repeated clip detection. In this letter, video sequences are represented as temporal trajectories via scaling and lower dimensional representation of the video frame luminance field, and a video trajectory indexing and matching scheme is developed to support video clip search. Simulation results demonstrate that the proposed approach achieves excellent performance in both response speed and precision-recall accuracy. © 2009 IEEE.

@article{Aggelos2009,
abstract = {With the fast-paced development of multimedia computing technologies, multimedia services have become pervasive. However, the design methodology of multimedia communications has still been based on the Internet architecture designed for loss-sensitive delay-insensitive data services, which is no longer applicable for multimedia communications. We believe user-perceived quality is the holy grail of multimedia communications systems. Thus, designing for quality will change the traditional paradigm of multimedia communications. Quality-driven cross-layer design focuses on bridging the gap between users' experience and multimedia systems, since it has been long ignored in previous research. We also hope this special issue can spark and stimulate the research and practice on quality-driven cross-layer optimized multimedia communications.},
author = {Katsaggelos, Aggelos K. and Ci, Song and Wang, Haohong and Zhang, Qian and Argyriou, Antonios},
doi = {10.1109/TMM.2009.2025681},
issn = {1520-9210},
journal = {IEEE Transactions on Multimedia},
month = {oct},
number = {6},
pages = {1049--1051},
title = {{Special Issue on Quality-Driven Cross-Layer Design for Multimedia Communications}},
url = {http://ieeexplore.ieee.org/document/5235193/},
volume = {11},
year = {2009}
}

With the fast-paced development of multimedia computing technologies, multimedia services have become pervasive. However, the design methodology of multimedia communications has still been based on the Internet architecture designed for loss-sensitive delay-insensitive data services, which is no longer applicable for multimedia communications. We believe user-perceived quality is the holy grail of multimedia communications systems. Thus, designing for quality will change the traditional paradigm of multimedia communications. Quality-driven cross-layer design focuses on bridging the gap between users' experience and multimedia systems, since it has been long ignored in previous research. We also hope this special issue can spark and stimulate the research and practice on quality-driven cross-layer optimized multimedia communications.

@article{Ehsan2009b,
author = {Ehsan, Maani and Aggelos, K Katsaggelos},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
pages = {407--416 , publisher = IEEE},
title = {{Unequal error protection for robust streaming of scalable video over packet lossy networks}},
volume = {20},
year = {2009}
}

@article{Ehsan2008,
abstract = {The newly adopted scalable extension of H.264/AVC video coding standard (SVC) demonstrates significant improvements in coding efficiency in addition to an increased degree of supported scalability relative to the scalable profiles of prior video coding standards. Due to the complicated hierarchical prediction structure of the SVC and the concept of key pictures, content-aware rate adaptation of SVC bit streams to intermediate bit rates is a nontrivial task. The concept of quality layers has been introduced in the design of the SVC to allow for fast content-aware prioritized rate adaptation. However, existing quality layer assignment methods are suboptimal and do not consider all network abstraction layer (NAL) units from different layers for the optimization. In this paper, we first propose a technique to accurately and efficiently estimate the quality degradation resulting from discarding an arbitrary number of NAL units from multiple layers of a bitstream by properly taking drift into account. Then, we utilize this distortion estimation technique to assign quality layers to NAL units for a more efficient extraction. Experimental results show that a significant gain can be achieved by the proposed scheme. {\textcopyright} 2009 IEEE.},
author = {Maani, Ehsan and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2009.2023152},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Automatic voltage control,Bit rate,Data mining,Encoding,No Index Terms Provided,Scalability,Spatial resolution,Static VAr compensators},
month = {sep},
number = {9},
pages = {2022--2029},
publisher = {IEEE},
title = {{Optimized Bit Extraction Using Distortion Modeling in the Scalable Extension of H.264/AVC}},
url = {http://ieeexplore.ieee.org/document/4926224/},
volume = {18},
year = {2009}
}

The newly adopted scalable extension of H.264/AVC video coding standard (SVC) demonstrates significant improvements in coding efficiency in addition to an increased degree of supported scalability relative to the scalable profiles of prior video coding standards. Due to the complicated hierarchical prediction structure of the SVC and the concept of key pictures, content-aware rate adaptation of SVC bit streams to intermediate bit rates is a nontrivial task. The concept of quality layers has been introduced in the design of the SVC to allow for fast content-aware prioritized rate adaptation. However, existing quality layer assignment methods are suboptimal and do not consider all network abstraction layer (NAL) units from different layers for the optimization. In this paper, we first propose a technique to accurately and efficiently estimate the quality degradation resulting from discarding an arbitrary number of NAL units from multiple layers of a bitstream by properly taking drift into account. Then, we utilize this distortion estimation technique to assign quality layers to NAL units for a more efficient extraction. Experimental results show that a significant gain can be achieved by the proposed scheme. © 2009 IEEE.

@article{wu2009quality,
abstract = {Service-oriented architecture provides a solution to the increasing network complexity due to ever-growing heterogeneous networks. As the most significant component of SOA, the decision engine is to create a workflow, defined as a sequence of individual data processing entities, for providing end-to-end QoS of a given task. Although the workflow of video transmission is generally known, existing solutions are often monolithic. Furthermore, there is no decision engine to select a workflow based on the best user-perceived quality. In this article we propose a service-oriented decision engine framework, which consists of a decision engine, a performance evaluation component, and other major SOA components to support real-time video transmission over wireless multihop networks, aiming to provide the best user-perceived video quality under application-centric QoS constraints. Based on the investigation of the stateof- the-art research efforts on SOA, some key issues for wireless live video transmission are discussed, and a case study for live video transmission is given to illustrate the proposed scheme. The superior performance of the proposed service-oriented decision engine is validated by experimental results. {\textcopyright} 2006 IEEE.},
author = {Wu, Dalei and Ci, Song and Luo, Haiyan and Wang, Haohong and Katsaggelos, Aggelos and {Dalei Wu} and {Song Ci} and {Haiyan Luo} and {Haohong Wang} and Katsaggelos, Aggelos},
doi = {10.1109/MWC.2009.5281255},
issn = {1536-1284},
journal = {IEEE Wireless Communications},
month = {aug},
number = {4},
pages = {48--54},
publisher = {IEEE},
title = {{A quality-driven decision engine for live video transmission under service-oriented architecture}},
url = {http://ieeexplore.ieee.org/document/5281255/},
volume = {16},
year = {2009}
}

Service-oriented architecture provides a solution to the increasing network complexity due to ever-growing heterogeneous networks. As the most significant component of SOA, the decision engine is to create a workflow, defined as a sequence of individual data processing entities, for providing end-to-end QoS of a given task. Although the workflow of video transmission is generally known, existing solutions are often monolithic. Furthermore, there is no decision engine to select a workflow based on the best user-perceived quality. In this article we propose a service-oriented decision engine framework, which consists of a decision engine, a performance evaluation component, and other major SOA components to support real-time video transmission over wireless multihop networks, aiming to provide the best user-perceived video quality under application-centric QoS constraints. Based on the investigation of the stateof- the-art research efforts on SOA, some key issues for wireless live video transmission are discussed, and a case study for live video transmission is given to illustrate the proposed scheme. The superior performance of the proposed service-oriented decision engine is validated by experimental results. © 2006 IEEE.

@article{Sotirios2009,
author = {Tsaftaris, S.A. and Katsaggelos, A.K.},
doi = {10.1109/TNB.2009.2026371},
issn = {1536-1241},
journal = {IEEE Transactions on NanoBioscience},
month = {sep},
number = {3},
pages = {259--270},
title = {{Retrieval Efficiency of DNA-Based Databases of Digital Signals}},
url = {http://ieeexplore.ieee.org/document/5161324/},
volume = {8},
year = {2009}
}

@article{dai2009softcuts,
abstract = {Designing effective image priors is of great interest to image super-resolution (SR), which is a severely under-determined problem. An edge smoothness prior is favored since it is able to suppress the jagged edge artifact effectively. However, for soft image edges with gradual intensity transitions, it is generally difficult to obtain analytical forms for evaluating their smoothness. This paper characterizes soft edge smoothness based on a novel SoftCuts metric by generalizing the Geocuts method [1]. The proposed soft edge smoothness measure can approximate the average length of all level lines in an intensity image. Thus, the total length of all level lines can be minimized effectively by integrating this new form of prior. In addition, this paper presents a novel combination of this soft edge smoothness prior and the alpha matting technique for color image SR, by adaptively normalizing image edges according to their $\alpha$-channel description. This leads to the adaptive SoftCuts algorithm, which represents a unified treatment of edges with different contrasts and scales. Experimental results are presented which demonstrate the effectiveness of the proposed method. {\textcopyright} 2009 IEEE.},
author = {{Shengyang Dai} and {Mei Han} and {Wei Xu} and {Ying Wu} and {Yihong Gong} and Katsaggelos, Aggelos K A.K. and Dai, Shengyang and Han, Mei and Xu, Wei and Wu, Ying and Gong, Yihong and Katsaggelos, Aggelos K A.K.},
doi = {10.1109/TIP.2009.2012908},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Edge smoothness,SoftCuts,Super-resolution (SR),$\alpha$-channel description},
month = {may},
number = {5},
pages = {969--981},
publisher = {IEEE},
title = {{SoftCuts: A Soft Edge Smoothness Prior for Color Image Super-Resolution}},
url = {http://ieeexplore.ieee.org/document/4808429/},
volume = {18},
year = {2009}
}

Designing effective image priors is of great interest to image super-resolution (SR), which is a severely under-determined problem. An edge smoothness prior is favored since it is able to suppress the jagged edge artifact effectively. However, for soft image edges with gradual intensity transitions, it is generally difficult to obtain analytical forms for evaluating their smoothness. This paper characterizes soft edge smoothness based on a novel SoftCuts metric by generalizing the Geocuts method [1]. The proposed soft edge smoothness measure can approximate the average length of all level lines in an intensity image. Thus, the total length of all level lines can be minimized effectively by integrating this new form of prior. In addition, this paper presents a novel combination of this soft edge smoothness prior and the alpha matting technique for color image SR, by adaptively normalizing image edges according to their $α$-channel description. This leads to the adaptive SoftCuts algorithm, which represents a unified treatment of edges with different contrasts and scales. Experimental results are presented which demonstrate the effectiveness of the proposed method. © 2009 IEEE.

@article{Martin2009,
abstract = {Image thresholding is a very common image processing operation, since almost all image processing schemes need some sort of separation of the pixels into different classes. In order to determine the thresholds, most methods analyze the histogram of the image. The optimal thresholds are often found by either minimizing or maximizing an objective function with respect to the values of the thresholds. By defining two classes of objective functions for which the optimal thresholds can be found by efficient algorithms, this paper provides a framework for determining the solution approach for current and future multilevel thresholding algorithms. We show, for example, that the method proposed by Otsu and other well-known methods have objective functions belonging to these classes. By implementing the algorithms in ANSI C and comparing their execution times, we can also make quantitative statements about their performance.},
author = {Eichmann, Marco},
doi = {10.1117/1.3073891},
issn = {1017-9909},
journal = {Journal of Electronic Imaging},
month = {jan},
number = {1},
pages = {013004},
title = {{Framework for efficient optimal multilevel image thresholding}},
url = {http://electronicimaging.spiedigitallibrary.org/article.aspx?doi=10.1117/1.3073891},
volume = {18},
year = {2009}
}

Image thresholding is a very common image processing operation, since almost all image processing schemes need some sort of separation of the pixels into different classes. In order to determine the thresholds, most methods analyze the histogram of the image. The optimal thresholds are often found by either minimizing or maximizing an objective function with respect to the values of the thresholds. By defining two classes of objective functions for which the optimal thresholds can be found by efficient algorithms, this paper provides a framework for determining the solution approach for current and future multilevel thresholding algorithms. We show, for example, that the method proposed by Otsu and other well-known methods have objective functions belonging to these classes. By implementing the algorithms in ANSI C and comparing their execution times, we can also make quantitative statements about their performance.

@article{Rafael2008,
author = {Molina, Rafael and Vega, Miguel and Mateos, Javier and Katsaggelos, Aggelos K.},
doi = {10.1016/j.acha.2007.03.006},
issn = {10635203},
journal = {Applied and Computational Harmonic Analysis},
month = {mar},
number = {2},
pages = {251--267},
title = {{Variational posterior distribution approximation in Bayesian super resolution reconstruction of multispectral images}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1063520307001133},
volume = {24},
year = {2008}
}

@article{Ioannis2008,
author = {Koktzoglou, Ioannis and Tsaftaris, Sotirios A and Li, Debiao and Katsaggelos, Aggelos K and Dharmakumar, Rohan},
doi = {10.1186/1532-429X-10-S1-A366},
issn = {1532-429X},
journal = {Journal of Cardiovascular Magnetic Resonance},
month = {oct},
number = {S1},
pages = {A366},
title = {{2097 Automated tracking of a passive intramyocardial needle with off-resonance mri: a feasibility study}},
url = {https://jcmr-online.biomedcentral.com/articles/10.1186/1532-429X-10-S1-A366},
volume = {10},
year = {2008}
}

@article{Stefanos2008,
abstract = {Super-resolution (SR) is the term used to define the process of estimating a high resolution (HR) image or a set of HR images from a set of low resolution (LR) observations. In this paper we propose a class of SR algorithms based on the maximum a posteriori (MAP) framework. These algorithms utilize a new multichannel image prior model, along with the state-of-the art image prior and observation models. Numerical experiments comparing the proposed algorithms, demonstrate the advantages of the adopted multichannel approach. copyright by EURASIP.},
author = {Stefanos, P Belekos and Nikolaos, P Galatsanos and Aggelos, K Katsaggelos and Belekos, Stefanos P and Galatsanos, Nikolaos P and Katsaggelos, Aggelos K},
doi = {10.1109/TIP.2010.2042115},
isbn = {22195491 , issue = 6},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {jun},
number = {6},
pages = {1451--1464},
title = {{Maximum a Posteriori Video Super-Resolution Using a New Multichannel Image Prior}},
url = {http://ieeexplore.ieee.org/document/5404316/},
volume = {19},
year = {2008}
}

Super-resolution (SR) is the term used to define the process of estimating a high resolution (HR) image or a set of HR images from a set of low resolution (LR) observations. In this paper we propose a class of SR algorithms based on the maximum a posteriori (MAP) framework. These algorithms utilize a new multichannel image prior model, along with the state-of-the art image prior and observation models. Numerical experiments comparing the proposed algorithms, demonstrate the advantages of the adopted multichannel approach. copyright by EURASIP.

@article{Aggelos2007e,
abstract = {In this paper we propose and analyze a globally and locally adaptive super-resolution Bayesian methodology for pansharpening of multispectral images. The methodology incorporates prior knowledge on the expected characteristics of the multispectral images uses the sensor characteristics to model the observation process of both panchromatic and multispectral images and includes information on the unknown parameters in the model in the form of hyperprior distributions. Using real and synthetic data, the pansharpened multispectral images are compared with the images obtained by other pansharpening methods and their quality is assessed both qualitatively and quantitatively. The Author 2008. Published by Oxford University Press on behalf of The British Computer Society. All rights reserved.},
author = {Vega, M. and Mateos, J. and Molina, R. and Katsaggelos, A.K.},
doi = {10.1093/comjnl/bxn031},
issn = {0010-4620},
journal = {The Computer Journal},
keywords = {Bayesian models,Hyperspectral images,Super-resolution},
month = {feb},
number = {1},
pages = {153--167},
title = {{Super-Resolution of Multispectral Images}},
url = {https://academic.oup.com/comjnl/article-lookup/doi/10.1093/comjnl/bxn031},
volume = {52},
year = {2008}
}

In this paper we propose and analyze a globally and locally adaptive super-resolution Bayesian methodology for pansharpening of multispectral images. The methodology incorporates prior knowledge on the expected characteristics of the multispectral images uses the sensor characteristics to model the observation process of both panchromatic and multispectral images and includes information on the unknown parameters in the model in the form of hyperprior distributions. Using real and synthetic data, the pansharpened multispectral images are compared with the images obtained by other pansharpening methods and their quality is assessed both qualitatively and quantitatively. The Author 2008. Published by Oxford University Press on behalf of The British Computer Society. All rights reserved.

@article{Jianwei2008,
abstract = {Multi-user video streaming over wireless channels is a challenging problem, where the demand for better video quality and small transmission delays needs to be reconciled with the limited and often time-varying communication resources. This paper presents a framework for joint network optimization, source adaptation, and deadline-driven scheduling for multi-user video streaming over wireless networks. We develop a joint adaptation, resource allocation and scheduling (JARS) algorithm, which allocates the communication resource based on the video users' quality of service, adapts video sources based on smart summarization, and schedules the transmissions to meet the frame delivery deadlines. The proposed algorithm leads to near full utilization of the network resources and satisfies the delivery deadlines for all video frames. Substantial performance improvements are achieved compared with heuristic schemes that do not take the interactions between multiple users into consideration. {\textcopyright} 2006 IEEE.},
author = {Jianwei, Huang and Zhu, Li and Mung, Chiang and Katsaggelos, A K},
chapter = {582},
doi = {10.1109/tcsvt.2008.919109},
isbn = {1051-8215
1558-2205},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
keywords = {Collaborative video streaming,Optimization decompo},
number = {5},
pages = {582--595},
title = {{Joint Source Adaptation and Resource Allocation for Multi-User Wireless Video Streaming}},
volume = {18},
year = {2008}
}

Multi-user video streaming over wireless channels is a challenging problem, where the demand for better video quality and small transmission delays needs to be reconciled with the limited and often time-varying communication resources. This paper presents a framework for joint network optimization, source adaptation, and deadline-driven scheduling for multi-user video streaming over wireless networks. We develop a joint adaptation, resource allocation and scheduling (JARS) algorithm, which allocates the communication resource based on the video users' quality of service, adapts video sources based on smart summarization, and schedules the transmissions to meet the frame delivery deadlines. The proposed algorithm leads to near full utilization of the network resources and satisfies the delivery deadlines for all video frames. Substantial performance improvements are achieved compared with heuristic schemes that do not take the interactions between multiple users into consideration. © 2006 IEEE.

@article{AggelosKonstantinos2010,
abstract = {The project has developed statistical approaches for the detection of abnormal video events for surveillance applications. The project proposes to extend such approaches and apply them towards the classification of vehicle trajectories in roadway video data for analysis and mitigation of traffic congestion. With the proposed approach, traffic information will first be analyzed off-line in an automated fashion. The project examines both the behavior of each vehicle independently but also its interaction with other vehicles. The effect of abnormal events onto incoming traffic will be a central objective of the investigation. The goal is to provide the foundations of a system that will allow for the off-line analysis of video data. The results of the off-line analysis could be utilized in two major ways: (i) by transportation officials to consider revising transportation rules and regulations and (ii) in developing on-line technologies for tracking the most disruptive abnormal events and minimizing their effect in creating congestion, via, for example, deployment of emergency vehicles, timely response of transportation agencies, and roadside information display systems.},
author = {K., Katsaggelos and S., Tsaftaris and Y., Wu and {Derin Babacan} and Jiang, Fan},
journal = {Transportation},
pages = {5},
title = {{Video Traffic Analysis for Abnormal Event Detection}},
url = {http://www.ccitt.northwestern.edu/documents/FinalReportvideotrafficanalysis.pdf},
year = {2008}
}

The project has developed statistical approaches for the detection of abnormal video events for surveillance applications. The project proposes to extend such approaches and apply them towards the classification of vehicle trajectories in roadway video data for analysis and mitigation of traffic congestion. With the proposed approach, traffic information will first be analyzed off-line in an automated fashion. The project examines both the behavior of each vehicle independently but also its interaction with other vehicles. The effect of abnormal events onto incoming traffic will be a central objective of the investigation. The goal is to provide the foundations of a system that will allow for the off-line analysis of video data. The results of the off-line analysis could be utilized in two major ways: (i) by transportation officials to consider revising transportation rules and regulations and (ii) in developing on-line technologies for tracking the most disruptive abnormal events and minimizing their effect in creating congestion, via, for example, deployment of emergency vehicles, timely response of transportation agencies, and roadside information display systems.

@article{babacan2008parameter,
abstract = {In this paper, we propose novel algorithms for total variation (TV) based image restoration and parameter estimation utilizing variational distribution approximations. Within the hierarchical Bayesian formulation, the reconstructed image and the unknown hyperparameters for the image prior and the noise are simultaneously estimated. The proposed algorithms provide approximations to the posterior distributions of the latent variables using variational methods. We show that some of the current approaches to TV-based image restoration are special cases of our framework. Experimental results show that the proposed approaches provide competitive performance without any assumptions about unknown hyperparameters and clearly outperform existing methods when additional information is included. {\textcopyright} 2008 IEEE.},
author = {Babacan, S.D. and Molina, Rafael and Katsaggelos, A.K.},
doi = {10.1109/TIP.2007.916051},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian methods,Image restoration,Parameter estimation,Total variation (TV),Variational methods},
month = {mar},
number = {3},
pages = {326--339},
pmid = {18270122},
publisher = {IEEE},
title = {{Parameter Estimation in TV Image Restoration Using Variational Distribution Approximation}},
url = {http://ieeexplore.ieee.org/document/4446214/},
volume = {17},
year = {2008}
}

In this paper, we propose novel algorithms for total variation (TV) based image restoration and parameter estimation utilizing variational distribution approximations. Within the hierarchical Bayesian formulation, the reconstructed image and the unknown hyperparameters for the image prior and the noise are simultaneously estimated. The proposed algorithms provide approximations to the posterior distributions of the latent variables using variational methods. We show that some of the current approaches to TV-based image restoration are special cases of our framework. Experimental results show that the proposed approaches provide competitive performance without any assumptions about unknown hyperparameters and clearly outperform existing methods when additional information is included. © 2008 IEEE.

@article{Konstantinos2005,
abstract = {We consider the transmission of a Gaussian source through a block fading channel. Assuming each block is decoded independently, the received distortion depends on the tradeoff between quantization accuracy and probability of outage. Namely, higher quantization accuracy requires a higher channel code rate, which increases the probability of outage. We first treat an outage as an erasure, and evaluate the received mean distortion with erasure coding across blocks as a function of the code length. We then evaluate the performance of scalable, or multi-resolution coding in which coded layers are superimposed within a coherence block, and the layers are sequentially decoded. Both the rate and power allocated to each layer are optimized. In addition to analyzing the performance with a finite number of layers, we evaluate the mean distortion at high Signal-to-Noise Ratios as the number of layers becomes infinite. As the block length of the erasure code increases to infinity, the received distortion converges to a deterministic limit, which is less than the mean distortion with an infinite-layer scalable coding scheme. However, for the same standard deviation in received distortion, infinite layer scalable coding performs slightly better than erasure coding, and with much less decoding delay. {\textcopyright} 2008 IEEE.},
author = {Zachariadis, Konstantinos and Honig, Michael and Katsaggelos, Aggelos},
doi = {10.1109/TCOMM.2008.060387},
issn = {0090-6778},
journal = {IEEE Transactions on Communications},
keywords = {Broadcast channel,Fading channel,Rate distortion,Scalable coding,Source-channel coding},
month = {jul},
number = {7},
pages = {1080--1091},
title = {{Source fidelity over fading channels: performance of erasure and scalable codes}},
url = {http://ieeexplore.ieee.org/document/4568449/},
volume = {56},
year = {2008}
}

We consider the transmission of a Gaussian source through a block fading channel. Assuming each block is decoded independently, the received distortion depends on the tradeoff between quantization accuracy and probability of outage. Namely, higher quantization accuracy requires a higher channel code rate, which increases the probability of outage. We first treat an outage as an erasure, and evaluate the received mean distortion with erasure coding across blocks as a function of the code length. We then evaluate the performance of scalable, or multi-resolution coding in which coded layers are superimposed within a coherence block, and the layers are sequentially decoded. Both the rate and power allocated to each layer are optimized. In addition to analyzing the performance with a finite number of layers, we evaluate the mean distortion at high Signal-to-Noise Ratios as the number of layers becomes infinite. As the block length of the erasure code increases to infinity, the received distortion converges to a deterministic limit, which is less than the mean distortion with an infinite-layer scalable coding scheme. However, for the same standard deviation in received distortion, infinite layer scalable coding performs slightly better than erasure coding, and with much less decoding delay. © 2008 IEEE.

@article{Yi2008,
abstract = {This paper addresses the problem of multi-user video transmission over the uplink of multi-carrier networks from an information theoretic perspective. Under the constraints imposed by the Physical (PHY) and Medium Access Control (MAC) layers, we exploit the unique property of state-of-the-art video coders that can provide bitstream prioritization in terms of distortion impact and solve the problem of allocating wireless resources, i.e., power and rate among multiple users such that the weighted sum of the overall video qualities is maximized. An optimality condition is derived to describe the achievable convex utility region. We start from the two-user case and develop an algorithm for the optimal resource allocation. Inspired by the intuition gained from the two-user case, we extend the algorithm to the multiple-user case. Our numerical simulations show that the proposed resource allocation algorithms give significant performance improvements as compared to application-layer agnostic solutions that do not consider the quality impact. {\textcopyright}2008 IEEE.},
author = {Maani, Ehsan and Pahalawatta, Peshala V. and Berry, Randall and Pappas, Thrasyvoulos N. and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2008.2001402},
isbn = {9781424420742 05361486 , issue = 9},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Multi-carrier networks,Utility driven resource man},
month = {sep},
number = {9},
pages = {1663--1671},
title = {{Resource Allocation for Downlink Multiuser Video Transmission Over Wireless Lossy Networks}},
url = {http://ieeexplore.ieee.org/document/4599190/},
volume = {17},
year = {2008}
}

This paper addresses the problem of multi-user video transmission over the uplink of multi-carrier networks from an information theoretic perspective. Under the constraints imposed by the Physical (PHY) and Medium Access Control (MAC) layers, we exploit the unique property of state-of-the-art video coders that can provide bitstream prioritization in terms of distortion impact and solve the problem of allocating wireless resources, i.e., power and rate among multiple users such that the weighted sum of the overall video qualities is maximized. An optimality condition is derived to describe the achievable convex utility region. We start from the two-user case and develop an algorithm for the optimal resource allocation. Inspired by the intuition gained from the two-user case, we extend the algorithm to the multiple-user case. Our numerical simulations show that the proposed resource allocation algorithms give significant performance improvements as compared to application-layer agnostic solutions that do not consider the quality impact. ©2008 IEEE.

@article{Stelios2008,
abstract = {This paper addresses the problem of data detection in orthogonal frequency division multiplexing (OFDM) systems operating under a time-varying multipath fading channel. Optimal detection in such a scenario is infeasible, which makes the introduction of approximations necessary. The typical joint data-channel estimators are decision directed, that is, assume perfect past data decisions. However, their performance is subject to error propagation phenomena. The variational Bayes method is employed here, which approximates the joint data and channel distribution as a separable one, greatly simplifying the problem. The data detection part of the resulting algorithm provides soft data estimates that are used for channel tracking. The channel itself is modeled as an autoregressive process allowing for a Kalman-like tracking algorithm. According to the developed algorithm, both data and channel estimates are exchanged and updated in an iterative manner. The performance of the proposed algorithm is evaluated by simulations. Furthermore, since OFDM is extremely sensitive to the presence of phase noise, the algorithm is extended to operate under severe phase noise conditions, with moderate performance degradation. {\textcopyright} 2008 IEEE.},
author = {Stefanatos, Stelios and Katsaggelos, A.K.},
doi = {10.1109/TSP.2008.925968},
issn = {1053-587X},
journal = {IEEE Transactions on Signal Processing},
keywords = {Channel tracking,Joint detection-estimation,Orthogonal frequency division multiplexing (OFDM),Phase noise,Variational Bayes method},
month = {sep},
number = {9},
pages = {4230--4243},
title = {{Joint Data Detection and Channel Tracking for OFDM Systems With Phase Noise}},
url = {http://ieeexplore.ieee.org/document/4527200/},
volume = {56},
year = {2008}
}

This paper addresses the problem of data detection in orthogonal frequency division multiplexing (OFDM) systems operating under a time-varying multipath fading channel. Optimal detection in such a scenario is infeasible, which makes the introduction of approximations necessary. The typical joint data-channel estimators are decision directed, that is, assume perfect past data decisions. However, their performance is subject to error propagation phenomena. The variational Bayes method is employed here, which approximates the joint data and channel distribution as a separable one, greatly simplifying the problem. The data detection part of the resulting algorithm provides soft data estimates that are used for channel tracking. The channel itself is modeled as an autoregressive process allowing for a Kalman-like tracking algorithm. According to the developed algorithm, both data and channel estimates are exchanged and updated in an iterative manner. The performance of the proposed algorithm is evaluated by simulations. Furthermore, since OFDM is extremely sensitive to the presence of phase noise, the algorithm is extended to operate under severe phase noise conditions, with moderate performance degradation. © 2008 IEEE.

@article{Javier2008,
abstract = {In this paper we propose a novel algorithm for the pansharpening of multispectral images based on the use of a Total Variation (TV) image prior. Within the Bayesian formulation, the proposed methodology incorporates prior knowledge on the expected characteristics of multispectral images, and uses the sensor characteristics to model the observation process of both panchromatic and multispectral images. The pansharpened multispectral images are compared with the images obtained by other parsharpening methods and their quality is assessed both qualitatively and quantitatively. {\textcopyright} 2008 IOP Publishing Ltd.},
author = {Mateos, Javier and Vega, Miguel and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1088/1742-6596/139/1/012022},
issn = {1742-6596},
journal = {Journal of Physics: Conference Series},
month = {nov},
pages = {012022},
title = {{Pansharpening of multispectral images using a TV-based super-resolution algorithm}},
url = {https://iopscience.iop.org/article/10.1088/1742-6596/139/1/012022},
volume = {139},
year = {2008}
}

In this paper we propose a novel algorithm for the pansharpening of multispectral images based on the use of a Total Variation (TV) image prior. Within the Bayesian formulation, the proposed methodology incorporates prior knowledge on the expected characteristics of multispectral images, and uses the sensor characteristics to model the observation process of both panchromatic and multispectral images. The pansharpened multispectral images are compared with the images obtained by other parsharpening methods and their quality is assessed both qualitatively and quantitatively. © 2008 IOP Publishing Ltd.

@article{Zhu2008,
abstract = {The deployment of the higher data rate wireless infrastructure systems and the emerging convergence of voice, video, and data services have been driving various modern multimedia applications, such as video streaming and mobile TV. However, the greatest challenge for video transmission over an uplink multiaccess wireless channel is the limited channel bandwidth and battery energy of a mobile device. In this paper, we pursue an energy-efficient video communication solution through joint video summarization and transmission adaptation over a slow fading wireless channel. Video summarization, coding and modulation schemes, and packet transmission are optimally adapted to the unique packet arrival and delay characteristics of the video summaries. In addition to the optimal solution, we also propose a heuristic solution that has close-to-optimal performance. Operational energy efficiency versus video distortion performance is characterized under a summarization setting. Simulation results demonstrate the advantage of the proposed scheme in energy efficiency and video transmission quality.},
author = {Li, Zhu and Zhai, Fan and Katsaggelos, Aggelos K.},
doi = {10.1155/2008/657032},
issn = {1687-6180},
journal = {EURASIP Journal on Advances in Signal Processing},
month = {dec},
number = {1},
pages = {657032},
title = {{Joint Video Summarization and Transmission Adaptation for Energy-Efficient Wireless Video Streaming}},
url = {https://asp-eurasipjournals.springeropen.com/articles/10.1155/2008/657032},
volume = {2008},
year = {2008}
}

The deployment of the higher data rate wireless infrastructure systems and the emerging convergence of voice, video, and data services have been driving various modern multimedia applications, such as video streaming and mobile TV. However, the greatest challenge for video transmission over an uplink multiaccess wireless channel is the limited channel bandwidth and battery energy of a mobile device. In this paper, we pursue an energy-efficient video communication solution through joint video summarization and transmission adaptation over a slow fading wireless channel. Video summarization, coding and modulation schemes, and packet transmission are optimally adapted to the unique packet arrival and delay characteristics of the video summaries. In addition to the optimal solution, we also propose a heuristic solution that has close-to-optimal performance. Operational energy efficiency versus video distortion performance is characterized under a summarization setting. Simulation results demonstrate the advantage of the proposed scheme in energy efficiency and video transmission quality.

@article{Yun2008,
abstract = {Subspace and similarity metric learning are important issues for image and video analysis in the scenarios of both computer vision and multimedia fields. Many real-world applications, such as image clustering/labeling and video indexing/retrieval, involve feature space dimensionality reduction as well as feature matching metric learning. However, the loss of information from dimensionality reduction may degrade the accuracy of similarity matching. In practice, such basic conflicting requirements for both feature representation efficiency and similarity matching accuracy need to be appropriately addressed. In the style of "Thinking Globally and Fitting Locally", we develop Locally Embedded Analysis (LEA) based solutions for visual data clustering and retrieval. LEA reveals the essential low-dimensional manifold structure of the data by preserving the local nearest neighbor affinity, and allowing a linear subspace embedding through solving a graph embedded eigenvalue decomposition problem. A visual data clustering algorithm, called Locally Embedded Clustering (LEC), and a local similarity metric learning algorithm for robust video retrieval, called Locally Adaptive Retrieval (LAR), are both designed upon the LEA approach, with variations in local affinity graph modeling. For large size database applications, instead of learning a global metric, we localize the metric learning space with kd-tree partition to localities identified by the indexing process. Simulation results demonstrate the effective performance of proposed solutions in both accuracy and speed aspects. {\textcopyright} 2007 Elsevier Inc. All rights reserved.},
author = {Fu, Yun and Li, Zhu and Huang, Thomas S. and Katsaggelos, Aggelos K.},
doi = {10.1016/j.cviu.2007.09.017},
issn = {10773142},
journal = {Computer Vision and Image Understanding},
keywords = {Dimensionality reduction,Image and video retrieval,Locally adaptive retrieval,Locally embedded analysis,Locally embedded clustering,Manifold,Similarity matching,Subspace learning,Visual clustering},
month = {jun},
number = {3},
pages = {390--402},
title = {{Locally adaptive subspace and similarity metric learning for visual data clustering and retrieval}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1077314207001518},
volume = {110},
year = {2008}
}

Subspace and similarity metric learning are important issues for image and video analysis in the scenarios of both computer vision and multimedia fields. Many real-world applications, such as image clustering/labeling and video indexing/retrieval, involve feature space dimensionality reduction as well as feature matching metric learning. However, the loss of information from dimensionality reduction may degrade the accuracy of similarity matching. In practice, such basic conflicting requirements for both feature representation efficiency and similarity matching accuracy need to be appropriately addressed. In the style of "Thinking Globally and Fitting Locally", we develop Locally Embedded Analysis (LEA) based solutions for visual data clustering and retrieval. LEA reveals the essential low-dimensional manifold structure of the data by preserving the local nearest neighbor affinity, and allowing a linear subspace embedding through solving a graph embedded eigenvalue decomposition problem. A visual data clustering algorithm, called Locally Embedded Clustering (LEC), and a local similarity metric learning algorithm for robust video retrieval, called Locally Adaptive Retrieval (LAR), are both designed upon the LEA approach, with variations in local affinity graph modeling. For large size database applications, instead of learning a global metric, we localize the metric learning space with kd-tree partition to localities identified by the indexing process. Simulation results demonstrate the effective performance of proposed solutions in both accuracy and speed aspects. © 2007 Elsevier Inc. All rights reserved.

@article{tsaftaris2008not,
author = {Tsaftaris, S. A. and Katsaggelos, A. K.},
doi = {10.1109/JPROC.2007.913496},
issn = {0018-9219},
journal = {Proceedings of the IEEE},
number = {3},
pages = {375--377},
publisher = {IEEE INSTITUTE OF ELECTRICAL AND ELECTRONICS},
title = {{The Not So Digital Future of Digital Signal Processing [Point of View]}},
url = {http://ieeexplore.ieee.org/document/4446225/},
volume = {96},
year = {2008}
}

@article{Fabrizio2007,
abstract = {Cross-layering is a design paradigm for overcoming the limitations deriving from the ISO/OSI layering principle, thus improving the performance of communications in specific scenarios, such as wireless multimedia communications. However, most available solutions are based on empirical considerations, and do not provide a theoretical background supporting such approaches. The paper aims at providing an analytical framework for the study of single-hop video delivery over a wireless link, enabling cross-layer interactions for performance optimization using power control and FEC and providing a useful tool to determine the potential gain deriving from the employment of such design paradigm. The analysis is performed using rate-distortion information of an embedded video bitstream jointly with a Lagrangian power minimization approach. Simulation results underline that cross-layering can provide relevant improvement in specific environments and that the proposed approach is able to capitalize on the advantage deriving from its deployment.},
author = {Granelli, Fabrizio and Costa, Cristina E. and Katsaggelos, Aggelos K.},
doi = {10.1155/2007/95807},
issn = {1687-5680},
journal = {Advances in Multimedia},
pages = {1--14},
title = {{A Study on the Usage of Cross-Layer Power Control and Forward Error Correction for Embedded Video Transmission over Wireless Links}},
url = {http://www.hindawi.com/journals/am/2007/095807/abs/},
volume = {2007},
year = {2007}
}

Cross-layering is a design paradigm for overcoming the limitations deriving from the ISO/OSI layering principle, thus improving the performance of communications in specific scenarios, such as wireless multimedia communications. However, most available solutions are based on empirical considerations, and do not provide a theoretical background supporting such approaches. The paper aims at providing an analytical framework for the study of single-hop video delivery over a wireless link, enabling cross-layer interactions for performance optimization using power control and FEC and providing a useful tool to determine the potential gain deriving from the employment of such design paradigm. The analysis is performed using rate-distortion information of an embedded video bitstream jointly with a Lagrangian power minimization approach. Simulation results underline that cross-layering can provide relevant improvement in specific environments and that the proposed approach is able to capitalize on the advantage deriving from its deployment.

@article{Barreto2007,
author = {Barreto, D. and Alvarez, L. D. and Molina, R. and Katsaggelos, A. K. and Callic{\'{o}}, G. M.},
doi = {10.1007/s11045-007-0019-y},
issn = {0923-6082},
journal = {Multidimensional Systems and Signal Processing},
month = {sep},
number = {2-3},
pages = {59--81},
title = {{Region-based super-resolution for compression}},
url = {https://link.springer.com/10.1007/s11045-007-0019-y},
volume = {18},
year = {2007}
}

@article{pahalawatta2007content,
abstract = {A cross-layer packet scheduling scheme that streams pre-encoded video over wireless downlink packet access networks to multiple users is presented. The scheme can be used with the emerging wireless standards such as HSDPA and IEEE 802.16. A gradient based scheduling scheme is used in which user data rates are dynamically adjusted based on channel quality as well as the gradients of a utility function. The user utilities are designed as a function of the distortion of the received video. This enables distortion-aware packet scheduling both within and across multiple users. The utility takes into account decoder error concealment, an important component in deciding the received quality of the video. We consider both simple and complex error concealment techniques. Simulation results show that the gradient based scheduling framework combined with the content-aware utility functions provides a viable method for downlink packet scheduling as it can significantly outperform current content-independent techniques. Further tests determine the sensitivity of the system to the initial video encoding schemes, as well as to non-real-time packet ordering techniques. {\textcopyright} 2007 IEEE.},
author = {Pahalawatta, Peshala and Berry, Randall and Pappas, Thrasyvoulos and Katsaggelos, Aggelos},
doi = {10.1109/JSAC.2007.070511},
issn = {0733-8716},
journal = {IEEE Journal on Selected Areas in Communications},
keywords = {Cross-layer design,H.264,HSDPA,Video streaming,Wireless packet scheduling},
month = {may},
number = {4},
pages = {749--759},
publisher = {IEEE},
title = {{Content-aware resource allocation and packet scheduling for video transmission over wireless networks}},
url = {http://ieeexplore.ieee.org/document/4205057/},
volume = {25},
year = {2007}
}

A cross-layer packet scheduling scheme that streams pre-encoded video over wireless downlink packet access networks to multiple users is presented. The scheme can be used with the emerging wireless standards such as HSDPA and IEEE 802.16. A gradient based scheduling scheme is used in which user data rates are dynamically adjusted based on channel quality as well as the gradients of a utility function. The user utilities are designed as a function of the distortion of the received video. This enables distortion-aware packet scheduling both within and across multiple users. The utility takes into account decoder error concealment, an important component in deciding the received quality of the video. We consider both simple and complex error concealment techniques. Simulation results show that the gradient based scheduling framework combined with the content-aware utility functions provides a viable method for downlink packet scheduling as it can significantly outperform current content-independent techniques. Further tests determine the sensitivity of the system to the initial video encoding schemes, as well as to non-real-time packet ordering techniques. © 2007 IEEE.

@article{de2007signal,
author = {{De Natale}, Francesco G. B. and Katsaggelos, Aggelos K. and Mayora, Oscar and Wu, Ying},
doi = {10.1155/2007/91730},
issn = {1687-6180},
journal = {EURASIP Journal on Advances in Signal Processing},
month = {dec},
number = {1},
pages = {091730},
publisher = {Springer International Publishing},
title = {{Signal Processing Technologies for Ambient Intelligence in Home-Care Applications}},
url = {https://asp-eurasipjournals.springeropen.com/articles/10.1155/2007/91730},
volume = {2007},
year = {2007}
}

@article{Peshala2007a,
abstract = {As wireless technology evolves towards its fourth generation (4G) of development, the prospect of offering multimedia services such as on-demand video streaming and video conferencing to wireless mobile clients becomes increasingly more viable. The eventual success of such applications depends on the efficient management of the limited system resources while taking into account the time-varying wireless channel conditions as well as the varying multimedia source content. In this paper, we review some of the recent advances in cross-layer design schemes, which aim at providing significant gains in performance for video streaming systems through content-aware resource allocation. Advances in both, real-time video streaming, where the video is encoded and transmitted in real-time, as well as, on-demand video streaming, where the video is pre-encoded in a media server, are considered. Copyright {\textcopyright} 2007 John Wiley & Sons, Ltd.},
author = {Pahalawatta, Peshala V. and Katsaggelos, Aggelos K.},
doi = {10.1002/wcm.469},
issn = {15308669},
journal = {Wireless Communications and Mobile Computing},
keywords = {Cross-layer optimization,Multi-user video transmission,Video streaming,Wireless video},
month = {feb},
number = {2},
pages = {131--142},
title = {{Review of content-aware resource allocation schemes for video streaming over wireless networks}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/wcm.469},
volume = {7},
year = {2007}
}

As wireless technology evolves towards its fourth generation (4G) of development, the prospect of offering multimedia services such as on-demand video streaming and video conferencing to wireless mobile clients becomes increasingly more viable. The eventual success of such applications depends on the efficient management of the limited system resources while taking into account the time-varying wireless channel conditions as well as the varying multimedia source content. In this paper, we review some of the recent advances in cross-layer design schemes, which aim at providing significant gains in performance for video streaming systems through content-aware resource allocation. Advances in both, real-time video streaming, where the video is encoded and transmitted in real-time, as well as, on-demand video streaming, where the video is pre-encoded in a media server, are considered. Copyright © 2007 John Wiley & Sons, Ltd.

@article{MinKyu2007,
author = {Park, Min Kyu and Kang, Moon-Gi and Katsaggelos, Aggelos K.},
doi = {10.1117/1.2802611},
issn = {0091-3286},
journal = {Optical Engineering},
month = {nov},
number = {11},
pages = {117004},
title = {{Regularized high-resolution image reconstruction considering inaccurate motion information}},
url = {http://opticalengineering.spiedigitallibrary.org/article.aspx?doi=10.1117/1.2802611},
volume = {46},
year = {2007}
}

@article{Miguel2006,
abstract = {Most of the available digital color cameras use a single image sensor with a color filter array (CFA) in acquiring an image. In order to produce a visible color image, a demosaicing process must be applied, which produces undesirable artifacts. An additional problem appears when the observed color image is also blurred. This paper addresses the problem of deconvolving color images observed with a single coupled charged device (CCD)from the super-resolution point of view. Utilizing the Bayesian paradigm, an estimate of the reconstructed image and the model parameters is generated. The proposed method is tested on real images. Copyright {\textcopyright} Hindawi Publishing Corporation. All rights reserved.},
author = {Vega, Miguel and Molina, Rafael and Katsaggelos, Aggelos K.},
doi = {10.1155/ASP/2006/25072},
issn = {1687-6180},
journal = {EURASIP Journal on Advances in Signal Processing},
month = {dec},
number = {1},
pages = {025072},
title = {{A Bayesian Super-Resolution Approach to Demosaicing of Blurred Images}},
url = {https://asp-eurasipjournals.springeropen.com/articles/10.1155/ASP/2006/25072},
volume = {2006},
year = {2006}
}

Most of the available digital color cameras use a single image sensor with a color filter array (CFA) in acquiring an image. In order to produce a visible color image, a demosaicing process must be applied, which produces undesirable artifacts. An additional problem appears when the observed color image is also blurred. This paper addresses the problem of deconvolving color images observed with a single coupled charged device (CCD)from the super-resolution point of view. Utilizing the Bayesian paradigm, an estimate of the reconstructed image and the model parameters is generated. The proposed method is tested on real images. Copyright © Hindawi Publishing Corporation. All rights reserved.

@article{Molina2006,
author = {Molina, Rafael and Mateos, Javier and Katsaggelos, Aggelos K.},
doi = {10.1109/TIP.2006.881972},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {dec},
number = {12},
pages = {3715--3727},
title = {{Blind Deconvolution Using a Variational Approach to Parameter, Image, and Blur Estimation}},
url = {http://ieeexplore.ieee.org/document/4011964/},
volume = {15},
year = {2006}
}

@article{wang2006joint,
abstract = {In recent years, joint source-channel coding for multimedia communications has gained increased popularity. However, very limited work has been conducted to address the problem of joint source-channel coding for object-based video. In this paper, we propose a data hiding scheme that improves the error resilience of object-based video by adaptively embedding the shape and motion information into the texture data. Within a rate-distortion theoretical framework, the source coding, channel coding, data embedding, and decoder error concealment are jointly optimized based on knowledge of the transmission channel conditions. Our goal is to achieve the best video quality as expressed by the minimum total expected distortion. The optimization problem is solved using Lagrangian relaxation and dynamic programming. The performance of the proposed scheme is tested using simulations of a Rayleigh-fading wireless channel, and the algorithm is implemented based on the MPEG-4 verification model. Experimental results indicate that the proposed hybrid source-channel coding scheme significantly outperforms methods without data hiding or unequal error protection. {\textcopyright} 2006 IEEE.},
author = {{Haohong Wang} and Tsaftaris, S.A. Sotirios A and Katsaggelos, Aggelos K A.K. and Wang, Haohong and Tsaftaris, S.A. Sotirios A and Katsaggelos, Aggelos K A.K.},
doi = {10.1109/TIP.2006.875194},
issn = {1057-7149},
journal = {IEEE Transactions on image processing},
keywords = {Data embedding,Data hiding,Joint source-channel coding,MPEG-4 standard,Object-based video,Rate-distortion,Unequal error protection (UEP),Video coding,Video communications,Wireless channel},
month = {aug},
number = {8},
pages = {2158--2169},
pmid = {16900673},
publisher = {IEEE},
title = {{Joint source-channel coding for wireless object-based video communications utilizing data hiding}},
url = {http://ieeexplore.ieee.org/document/1658082/},
volume = {15},
year = {2006}
}

In recent years, joint source-channel coding for multimedia communications has gained increased popularity. However, very limited work has been conducted to address the problem of joint source-channel coding for object-based video. In this paper, we propose a data hiding scheme that improves the error resilience of object-based video by adaptively embedding the shape and motion information into the texture data. Within a rate-distortion theoretical framework, the source coding, channel coding, data embedding, and decoder error concealment are jointly optimized based on knowledge of the transmission channel conditions. Our goal is to achieve the best video quality as expressed by the minimum total expected distortion. The optimization problem is solved using Lagrangian relaxation and dynamic programming. The performance of the proposed scheme is tested using simulations of a Rayleigh-fading wireless channel, and the algorithm is implemented based on the MPEG-4 verification model. Experimental results indicate that the proposed hybrid source-channel coding scheme significantly outperforms methods without data hiding or unequal error protection. © 2006 IEEE.

@article{schuster2004motion,
abstract = {The introduction of Video Objects (VOs) is one of the innovations of MPEG-4. The $\alpha$plane of a VO defines its shape at a given instance in time and hence determines the boundary of its texture. In packet-based networks, shape, motion, and texture are subject to loss. While there has been considerable attention paid to the concealment of texture and motion errors, little has been done in the field of shape error concealment. In this paper we propose a post-processing shape err or concealment technique that uses the motion compensated boundary information of the previously received $\alpha$-plane. The proposed approach is based on matching received boundary segments in the current frame to the boundary in the previous frame. This matching is achieved by finding a maximally smooth motion vector field. After the current boundary segments are matched to the previous boundary, the missing boundary pieces are reconstructed by motion compensation. Experimental results demonstrating the performance of the proposed motion compensated shape error concealment method, and comparing it with the previously proposed weighted side matching method [1] are presented. {\textcopyright} 2006 IEEE.},
author = {Schuster, G.M. and Katsaggelos, A.K.},
doi = {10.1109/ICIP.2004.1418718},
institution = {IEEE},
isbn = {0-7803-8554-3},
issn = {1057-7149},
journal = {2004 International Conference on Image Processing, 2004. ICIP '04.},
keywords = {Dynamic programming,Error concealment,MPEG-4,Motion compensation},
month = {feb},
number = {2},
pages = {501--510},
pmid = {16479820},
publisher = {IEEE},
title = {{Motion compensated shape error concealment}},
url = {https://ieeexplore.ieee.org/document/1576823 http://ieeexplore.ieee.org/document/1418718/},
volume = {1},
year = {2006}
}

The introduction of Video Objects (VOs) is one of the innovations of MPEG-4. The $α$plane of a VO defines its shape at a given instance in time and hence determines the boundary of its texture. In packet-based networks, shape, motion, and texture are subject to loss. While there has been considerable attention paid to the concealment of texture and motion errors, little has been done in the field of shape error concealment. In this paper we propose a post-processing shape err or concealment technique that uses the motion compensated boundary information of the previously received $α$-plane. The proposed approach is based on matching received boundary segments in the current frame to the boundary in the previous frame. This matching is achieved by finding a maximally smooth motion vector field. After the current boundary segments are matched to the previous boundary, the missing boundary pieces are reconstructed by motion compensation. Experimental results demonstrating the performance of the proposed motion compensated shape error concealment method, and comparing it with the previously proposed weighted side matching method [1] are presented. © 2006 IEEE.

@article{aleksic2006automatic,
abstract = {The performance of an automatic facial expression recognition system can be significantly improved by modeling the reliability of different streams of facial expression information utilizing multistream hidden Markov models (HMMs). In this paper, we present an automatic multistream HMM facial expression recognition system and analyze its performance. The proposed system utilizes facial animation parameters (FAPs), supported by the MPEG-4 standard, as features for facial expression classification. Specifically, the FAPs describing the movement of the outer-lip contours and eyebrows are used as observations. Experiments are first performed employing single-stream HMMs under several different scenarios, utilizing outer-lip and eyebrow FAPs individually and jointly. A multistream HMM approach is proposed for introducing facial expression and FAP group dependent stream reliability weights. The stream weights are determined based on the facial expression recognition results obtained when FAP streams are utilized individually. The proposed multistream HMM facial expression system, which utilizes stream reliability weights, achieves relative reduction of the facial expression recognition error of 44% compared to the single-stream HMM system. Copyright 2006, IEE.},
author = {Aleksic, P.S. and Katsaggelos, A.K.},
doi = {10.1109/TIFS.2005.863510},
issn = {1556-6013},
journal = {IEEE Transactions on Information Forensics and Security},
keywords = {Automatic facial expression recognition,Facial animation parameters,Hidden Markov models,MPEG-4 standards,Multistream HMM},
month = {mar},
number = {1},
pages = {3--11},
publisher = {IEEE},
title = {{Automatic Facial Expression Recognition Using Facial Animation Parameters and Multistream HMMs}},
url = {http://ieeexplore.ieee.org/document/1597130/},
volume = {1},
year = {2006}
}

The performance of an automatic facial expression recognition system can be significantly improved by modeling the reliability of different streams of facial expression information utilizing multistream hidden Markov models (HMMs). In this paper, we present an automatic multistream HMM facial expression recognition system and analyze its performance. The proposed system utilizes facial animation parameters (FAPs), supported by the MPEG-4 standard, as features for facial expression classification. Specifically, the FAPs describing the movement of the outer-lip contours and eyebrows are used as observations. Experiments are first performed employing single-stream HMMs under several different scenarios, utilizing outer-lip and eyebrow FAPs individually and jointly. A multistream HMM approach is proposed for introducing facial expression and FAP group dependent stream reliability weights. The stream weights are determined based on the facial expression recognition results obtained when FAP streams are utilized individually. The proposed multistream HMM facial expression system, which utilizes stream reliability weights, achieves relative reduction of the facial expression recognition error of 44% compared to the single-stream HMM system. Copyright 2006, IEE.

@article{Petar2006,
author = {Aleksic, P.S. and Katsaggelos, A.K.},
doi = {10.1109/JPROC.2006.886017},
issn = {0018-9219},
journal = {Proceedings of the IEEE},
month = {nov},
number = {11},
pages = {2025--2044},
title = {{Audio-Visual Biometrics}},
url = {https://ieeexplore.ieee.org/iel5/5/4052463/04052464.pdf http://ieeexplore.ieee.org/document/4052464/},
volume = {94},
year = {2006}
}

@article{Zhu2006a,
author = {Li, Zhu and Huang, Jianwei and Katsaggelos, Aggelos K and Chiang, Mung},
journal = {China Communications},
keywords = {collaborative video,optimization decomposition,pricing control,rate-distortion modeling,streaming,video adaptation},
number = {October},
pages = {58--70},
title = {{Intelligent Wireless Video Communication : Source Adaptation and Multi-User Collaboration}},
url = {https://www.academia.edu/download/31006241/7.pdf},
year = {2006}
}

@article{Jian-wei2006,
abstract = {We solve the problem of uplink video streaming in CDMA cellular networks by jointly designing the rate control and scheduling algorithms. In the pricing-based distributed rate control algorithm, the base station announces a price for the per unit average rate it can support, and the mobile devices choose their desired average transmission rates by balancing their video quality and cost of transmission. Each mobile device then determines the specific video frames to transmit by a video summarization process. In the time-division-multiplexing (TDM) scheduling algorithm, the base station collects the information on frames to be transmitted from all devices within the current time window, sorts them in increasing order of deadlines, and schedules the transmissions in a TDM fashion. This joint algorithm takes advantage of the multi-user content diversity, and maximizes the network total utility (i.e., minimize the network total distortion), while satisfying the delivery deadline constraints. Simulations showed that the proposed algorithm significantly outperforms the constant rate provision algorithm.},
author = {Huang, Jian-wei and Li, Zhu and Chiang, Mung and Katsaggelos, Aggelos K.},
doi = {10.1631/jzus.2006.A0801},
issn = {1673-565X},
journal = {Journal of Zhejiang University-SCIENCE A},
keywords = {Code division multiple access,Cross-layer design,Pricing,Uplink communications,Video streaming,Video summarization},
month = {may},
number = {5},
pages = {801--810},
title = {{Joint rate control and scheduling for wireless uplink video streaming}},
url = {http://link.springer.com/10.1631/jzus.2006.A0801},
volume = {7},
year = {2006}
}

We solve the problem of uplink video streaming in CDMA cellular networks by jointly designing the rate control and scheduling algorithms. In the pricing-based distributed rate control algorithm, the base station announces a price for the per unit average rate it can support, and the mobile devices choose their desired average transmission rates by balancing their video quality and cost of transmission. Each mobile device then determines the specific video frames to transmit by a video summarization process. In the time-division-multiplexing (TDM) scheduling algorithm, the base station collects the information on frames to be transmitted from all devices within the current time window, sorts them in increasing order of deadlines, and schedules the transmissions in a TDM fashion. This joint algorithm takes advantage of the multi-user content diversity, and maximizes the network total utility (i.e., minimize the network total distortion), while satisfying the delivery deadline constraints. Simulations showed that the proposed algorithm significantly outperforms the constant rate provision algorithm.

@article{woods2005stochastic,
abstract = {Using a stochastic framework, we propose two algorithms for the problem of obtaining a single high-resolution image from multiple noisy, blurred, and undersampled images. The first is based on a Bayesian formulation that is implemented via the expectation maximization algorithm. The second is based on a maximum a posteriori formulation. In both of our formulations, the registration, noise, and image statistics are treated as unknown parameters. These unknown parameters and the high-resolution image are estimated jointly based on the available observations. We present an efficient implementation of these algorithms in the frequency domain that allows their application to large images. Simulations are presented that test and compare the proposed algorithms. {\textcopyright} 2006 IEEE.},
author = {Woods, N.A. and Galatsanos, N.P. and Katsaggelos, A.K.},
doi = {10.1109/TIP.2005.860355},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {jan},
number = {1},
pages = {201--213},
pmid = {16435550},
publisher = {IEEE},
title = {{Stochastic methods for joint registration, restoration, and interpolation of multiple undersampled images}},
url = {https://ieeexplore.ieee.org/document/1556638},
volume = {15},
year = {2006}
}

Using a stochastic framework, we propose two algorithms for the problem of obtaining a single high-resolution image from multiple noisy, blurred, and undersampled images. The first is based on a Bayesian formulation that is implemented via the expectation maximization algorithm. The second is based on a maximum a posteriori formulation. In both of our formulations, the registration, noise, and image statistics are treated as unknown parameters. These unknown parameters and the high-resolution image are estimated jointly based on the available observations. We present an efficient implementation of these algorithms in the frequency domain that allows their application to large images. Simulations are presented that test and compare the proposed algorithms. © 2006 IEEE.

@article{katsaggelos2006efficient,
author = {Katsaggelos, A K and Pahalawatta, P V},
journal = {Virtual Observatory: Plate Content Digitization, Archive Mining and Image Sequence Processing},
pages = {234--244},
title = {{Efficient Video Communication over Lossy Channels}},
url = {https://ui.adsabs.harvard.edu/abs/2006vopc.conf..234K},
year = {2006}
}

@article{Yiftach2006,
author = {Eisenberg, Yiftach and {Fan Zhai} and Pappas, T.N. and Berry, Randall and Katsaggelos, A.K.},
doi = {10.1109/TIP.2005.860600},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {feb},
number = {2},
pages = {289--299},
title = {{VAPOR: variance-aware per-pixel optimal resource allocation}},
url = {https://ieeexplore.ieee.org/document/1576802},
volume = {15},
year = {2006}
}

@article{wang2005rate,
abstract = {In object-based video encoding, the encoding of the video data is decoupled into the encoding of shape, motion, and texture information, which enables certain functionalities, like content-based interactivity and content-based scalability. The fundamental problem, however, of how to jointly encode this separate information to reach the best coding efficiency has not been studied thoroughly. In this paper, we present an operational rate-distortion optimal scheme for the allocation of bits among shape, motion, and texture in object-based video encoding. Our approach is based on Lagrangian relaxation and dynamic programming. We implement our algorithm on the MPEG-4 video verification model, although it is applicable to any object-based video encoding scheme. The performance is accessed utilizing a proposed metric that jointly captures the distortion due to the encoding of the shape and texture. Experimental results demonstrate that the gains of lossy shape encoding depend on the percentage the shape bits occupy out of the total bit budget. This gain may be small or may be realized at very low bit rates for certain typical scenes. {\textcopyright} 2005 IEEE.},
author = {Wang, Haohong and Schuster, G.M. Guido M. and Katsaggelos, Aggelos K. A.K. and {Haohong Wang} and Schuster, G.M. Guido M. and Katsaggelos, Aggelos K. A.K.},
chapter = {1113},
doi = {10.1109/TCSVT.2005.852629},
isbn = {1051-8215},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
keywords = {MPEG-4,Object-based video,Rate-distortion,Shape coding,Video coding},
month = {sep},
number = {9},
pages = {1113--1123},
publisher = {IEEE},
title = {{Rate-distortion optimal bit allocation for object-based video coding}},
url = {http://ieeexplore.ieee.org/document/1501879/},
volume = {15},
year = {2005}
}

In object-based video encoding, the encoding of the video data is decoupled into the encoding of shape, motion, and texture information, which enables certain functionalities, like content-based interactivity and content-based scalability. The fundamental problem, however, of how to jointly encode this separate information to reach the best coding efficiency has not been studied thoroughly. In this paper, we present an operational rate-distortion optimal scheme for the allocation of bits among shape, motion, and texture in object-based video encoding. Our approach is based on Lagrangian relaxation and dynamic programming. We implement our algorithm on the MPEG-4 video verification model, although it is applicable to any object-based video encoding scheme. The performance is accessed utilizing a proposed metric that jointly captures the distortion due to the encoding of the shape and texture. Experimental results demonstrate that the gains of lossy shape encoding depend on the percentage the shape bits occupy out of the total bit budget. This gain may be small or may be realized at very low bit rates for certain typical scenes. © 2005 IEEE.

@article{lopezbayesian,
abstract = {ABSTRACT Emission tomography images are degraded due to the presence of noise and several physical factors like attenuation and scattering. To remove the attenuation effect from the emission tomography reconstruction, attenuation correction factors (ACFs) are used. These ACFs are obtained from a transmission scan and it is well known that they are homogeneous within each tissue and present abrupt variations in the transition between tissues.},
author = {L{\'{o}}pez, A and Garrido, J and Molina, R and Katsaggelos, A K},
journal = {ICIP},
title = {{Bayesian Transmission Image Reconstruction Using Compound Gauss-Markov Prior Models}},
url = {https://www.academia.edu/2704617/BAYESIAN_TRANSMISSION_IMAGE_RECONSTRUCTION_USING_COMPOUND_GAUSS_MARKOV_PRIOR_MODELS},
year = {2005}
}

ABSTRACT Emission tomography images are degraded due to the presence of noise and several physical factors like attenuation and scattering. To remove the attenuation effect from the emission tomography reconstruction, attenuation correction factors (ACFs) are used. These ACFs are obtained from a transmission scan and it is well known that they are homogeneous within each tissue and present abrupt variations in the transition between tissues.

@article{zhai2005joint,
abstract = {Differentiated Services (DiffServ) is one of the leading architectures for providing quality of service in the Internet. We propose a scheme for real-time video transmission over a DiffServ network that jointly considers video source coding, packet classification, and error concealment within a framework of costdistortion optimization. The selections of encoding parameters and packet classification are both used to manage end-to-end delay variations and packet losses within the network. We present two dual formulations of the proposed scheme: the minimum distortion problem, in which the objective is to minimize the end-to-end distortion subject to cost and delay constraints, and the minimum cost problem, which minimizes the total cost subject to end-to-end distortion and delay constraints. A solution to these problems using Lagrangian relaxation and dynamic programming is given. Simulation results demonstrate the advantage of jointly adapting the source coding and packet classification in DiffServ networks. {\textcopyright} 2005 IEEE.},
author = {Zhai, Fan and Luna, Carlos E. C.E. and Eisenberg, Yiftach and Pappas, T.N. Thrasyvoulos N. and Berry, Randall and Katsaggelos, Aggelos K. A.K. and {Fan Zhai} and Luna, Carlos E. C.E. and Eisenberg, Yiftach and Pappas, T.N. Thrasyvoulos N. and Berry, Randall and Katsaggelos, Aggelos K. A.K.},
chapter = {716},
doi = {10.1109/TMM.2005.850989},
isbn = {1520-9210},
issn = {15209210},
journal = {IEEE Transactions on Multimedia},
keywords = {Error concealment,Error resilience,Joint source-ch,Joint source-channel coding,Multimedia communication,Optimal resource allocation,QoS,Unequal error protection (UEP)},
month = {aug},
number = {4},
pages = {716--725},
publisher = {IEEE},
title = {{Joint source coding and packet classification for real-time video transmission over Differentiated Services networks}},
url = {http://ieeexplore.ieee.org/document/1468156/},
volume = {7},
year = {2005}
}

Differentiated Services (DiffServ) is one of the leading architectures for providing quality of service in the Internet. We propose a scheme for real-time video transmission over a DiffServ network that jointly considers video source coding, packet classification, and error concealment within a framework of costdistortion optimization. The selections of encoding parameters and packet classification are both used to manage end-to-end delay variations and packet losses within the network. We present two dual formulations of the proposed scheme: the minimum distortion problem, in which the objective is to minimize the end-to-end distortion subject to cost and delay constraints, and the minimum cost problem, which minimizes the total cost subject to end-to-end distortion and delay constraints. A solution to these problems using Lagrangian relaxation and dynamic programming is given. Simulation results demonstrate the advantage of jointly adapting the source coding and packet classification in DiffServ networks. © 2005 IEEE.

@article{Fan2005a,
author = {Fan, Zhai and Yiftach, Eisenberg and Aggelos, K Katsaggelos},
journal = {Handbook of Image and Video Processing},
pages = {1065--1082 , publisher = Academic},
title = {{Joint source-channel coding for video communications}},
url = {http://www.eecs.northwestern.edu/$\sim$fzhai/publications/Handbook_Image_Video_Processing_bookchapter.pdf},
year = {2005}
}

@article{wang2005cost,
abstract = {Object-based video coding is a relatively new technique to meet the fast growing demand for interactive multimedia applications. Compared with conventional frame-based video coding, it consists of two types of source data: shape information and texture information. Recently, joint source-channel coding for multimedia communications has gained increased popularity. However, very limited work has been conducted to address the problem of joint source-channel coding for object-based video. In this paper, we propose a cost-distortion optimal unequal error protection (UEP) scheme for object-based video communications. Our goal is to achieve the best video quality (minimum total expected distortion) with constraints on transmission cost and delay in a lossy network environment. The problem is solved using Lagarangian relaxation and dynamic programming. The performance of the proposed scheme is tested using simulations of a narrow-band block-fading wireless channel with additive white Gaussian noise and a simplified differentiated services Internet channel. Experimental results indicate that the proposed UEP scheme can significantly outperform equal error protection methods. {\textcopyright} 2005 IEEE.},
author = {{Haohong Wang} and {Fan Zhai} and Eisenberg, Yiftach and Katsaggelos, Aggelos K A.K. and Wang, Haohong and Zhai, Fan and Eisenberg, Yiftach and Katsaggelos, Aggelos K A.K.},
doi = {10.1109/TCSVT.2005.857305},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
keywords = {Differentiated service (Diffserv) network,Joint source-channel coding,Lossy network,MPEG-4 standard,Object-based video,Rate distortion,Unequal error protection (UEP),Video coding,Video communications,Wireless channel},
month = {dec},
number = {12},
pages = {1505--1516},
publisher = {IEEE},
title = {{Cost-distortion optimized unequal error protection for object-based video communications}},
url = {http://ieeexplore.ieee.org/document/1545999/},
volume = {15},
year = {2005}
}

Object-based video coding is a relatively new technique to meet the fast growing demand for interactive multimedia applications. Compared with conventional frame-based video coding, it consists of two types of source data: shape information and texture information. Recently, joint source-channel coding for multimedia communications has gained increased popularity. However, very limited work has been conducted to address the problem of joint source-channel coding for object-based video. In this paper, we propose a cost-distortion optimal unequal error protection (UEP) scheme for object-based video communications. Our goal is to achieve the best video quality (minimum total expected distortion) with constraints on transmission cost and delay in a lossy network environment. The problem is solved using Lagarangian relaxation and dynamic programming. The performance of the proposed scheme is tested using simulations of a narrow-band block-fading wireless channel with additive white Gaussian noise and a simplified differentiated services Internet channel. Experimental results indicate that the proposed UEP scheme can significantly outperform equal error protection methods. © 2005 IEEE.

@article{Tzavidas2005,
author = {Tzavidas, S. and Katsaggelos, A.K.},
doi = {10.1109/TMM.2005.854430},
issn = {1520-9210},
journal = {IEEE Transactions on Multimedia},
month = {oct},
number = {5},
pages = {880--890},
title = {{A multicamera setup for generating stereo panoramic video}},
url = {http://ieeexplore.ieee.org/document/1510635/},
volume = {7},
year = {2005}
}

@article{li2005minmax,
abstract = {The need for video summarization originates primarily from a viewing time constraint. A shorter version of the original video sequence is desirable in a number of applications. Clearly, a shorter version is also necessary in applications where storage, communication bandwidth and/or power are limited. In this paper, our work is based on a MINMAX optimization formulation with viewing time, frame skip and bit rate constraints. New metrics for missing frame and video summary distortions are introduced. Optimal algorithm based on dynamic programming is presented along with experimental results. {\textcopyright} 2005 IEEE.},
author = {{Zhu Li} and Schuster, G.M. Guido M. and Katsaggelos, Aggelos K. A.K. and Li, Zhu and Schuster, G.M. Guido M. and Katsaggelos, Aggelos K. A.K.},
chapter = {1245},
doi = {10.1109/TCSVT.2005.854230},
isbn = {1051-8215},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
keywords = {Dynamic programming,Rate-distortion optimization,Video analysis,Video summarization},
month = {oct},
number = {10},
pages = {1245--1256},
publisher = {IEEE},
title = {{MINMAX optimal video summarization}},
url = {http://ieeexplore.ieee.org/document/1512242/},
volume = {15},
year = {2005}
}

The need for video summarization originates primarily from a viewing time constraint. A shorter version of the original video sequence is desirable in a number of applications. Clearly, a shorter version is also necessary in applications where storage, communication bandwidth and/or power are limited. In this paper, our work is based on a MINMAX optimization formulation with viewing time, frame skip and bit rate constraints. New metrics for missing frame and video summary distortions are introduced. Optimal algorithm based on dynamic programming is presented along with experimental results. © 2005 IEEE.

@article{Sotirios2005,
author = {Tsaftaris, Sotirios and Hatzimanikatis, Vassily and Katsaggelos, Aggelos},
journal = {Journal of the Japan Society of Simulation Technology (JSST) special issue "Application and Simulation of DNA Computing"},
keywords = {database,dna,dna_dsp,tsaftaris},
number = {4},
pages = {268--276},
title = {{In silico estimation of annealing specificity of query searches in DNA databases}},
volume = {24},
year = {2005}
}

@article{li2005fast,
abstract = {Content-based video retrieval technology holds the key to the efficient management and sharing of video content from different sources, in different scales, across different platforms, and over different communication channels. In this work fast video shot retrieval algorithms based on the geometry of video sequence traces in the principal component space are presented. Techniques to address scale (spatial and temporal) issues, in addition to noise and other possible distortions, such as frame dropping, are discussed. Experimental results demonstrate the eifectiveness of the proposed approach. {\textcopyright} IEE, 2005.},
author = {Li, Z. and Katsaggelos, A.K. and Gandhi, B.},
doi = {10.1049/ip-vis:20045083},
issn = {1350245X},
journal = {IEE Proceedings - Vision, Image, and Signal Processing},
number = {3},
pages = {367},
publisher = {IET Digital Library},
title = {{Fast video shot retrieval based on trace geometry matching}},
url = {https://digital-library.theiet.org/content/journals/10.1049/ip-vis_20045083},
volume = {152},
year = {2005}
}

Content-based video retrieval technology holds the key to the efficient management and sharing of video content from different sources, in different scales, across different platforms, and over different communication channels. In this work fast video shot retrieval algorithms based on the geometry of video sequence traces in the principal component space are presented. Techniques to address scale (spatial and temporal) issues, in addition to noise and other possible distortions, such as frame dropping, are discussed. Experimental results demonstrate the eifectiveness of the proposed approach. © IEE, 2005.

@article{Hong2005,
author = {Hong, Min-Cheol and Stathaki, Tania and Katsaggelos, Aggelos K.},
doi = {10.1117/1.1867452},
issn = {1017-9909},
journal = {Journal of Electronic Imaging},
month = {jan},
number = {1},
pages = {013004},
title = {{Iterative regularized mixed norm multichannel image restoration}},
url = {http://electronicimaging.spiedigitallibrary.org/article.aspx?doi=10.1117/1.1867452},
volume = {14},
year = {2005}
}

@article{zhai2003joint,
abstract = {We consider efficiently transmitting video over a hybrid wireless/wire-line network by optimally allocating resources across multiple protocol layers. Specifically, we present a framework of joint source-channel coding and power adaptation, where error resilient source coding, channel coding, and transmission power adaptation are jointly designed to optimize video quality given constraints on the total transmission energy and delay for each video frame. In particular, we consider the combination of two types of channel coding - inter-packet coding (at the transport layer) to provide protection against packet dropping in the wire-line network and intra-packet coding (at the link layer) to provide protection against bit errors in the wireless link. In both cases, we allow the coding rate to be adaptive to provide unequal error protection at both the packet and frame level. In addition to both types of channel coding, we also compensate for channel errors by adapting the transmission power used to send each packet. An efficient algorithm based on Lagrangian relaxation and the method of alternating variables is proposed to solve the resulting optimization problem. Simulation results are shown to illustrate the advantages of joint optimization across multiple layers. {\textcopyright} 2005 Elsevier B.V. All rights reserved.},
author = {Zhai, Fan and Eisenberg, Yiftach and Pappas, Thrasyvoulos N. and Berry, Randall and Katsaggelos, Aggelos K.},
doi = {10.1016/j.image.2005.02.002},
institution = {The University; 1998},
issn = {09235965},
journal = {Signal Processing: Image Communication},
keywords = {Cross layer,Error concealment,Error resilience,Multimedia streaming,Product FEC,QoS,Unequal error protection (UEP)},
month = {apr},
number = {4},
pages = {371--387},
title = {{Joint source-channel coding and power adaptation for energy efficient wireless video communications}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0923596505000123},
volume = {20},
year = {2005}
}

We consider efficiently transmitting video over a hybrid wireless/wire-line network by optimally allocating resources across multiple protocol layers. Specifically, we present a framework of joint source-channel coding and power adaptation, where error resilient source coding, channel coding, and transmission power adaptation are jointly designed to optimize video quality given constraints on the total transmission energy and delay for each video frame. In particular, we consider the combination of two types of channel coding - inter-packet coding (at the transport layer) to provide protection against packet dropping in the wire-line network and intra-packet coding (at the link layer) to provide protection against bit errors in the wireless link. In both cases, we allow the coding rate to be adaptive to provide unequal error protection at both the packet and frame level. In addition to both types of channel coding, we also compensate for channel errors by adapting the transmission power used to send each packet. An efficient algorithm based on Lagrangian relaxation and the method of alternating variables is proposed to solve the resulting optimization problem. Simulation results are shown to illustrate the advantages of joint optimization across multiple layers. © 2005 Elsevier B.V. All rights reserved.

@article{molina2005super,
author = {Molina, R and Mateos, J and Katsaggelos, A K},
journal = {Virtual observatories: Plate Content Digitization, Archive Mining and Image Sequence processing, edited by Henron Press, Sofia (Bulgary)},
pages = {1--10},
title = {{Super resolution reconstruction of multispectral images ∗}},
url = {https://ccia.ugr.es/vip/files/conferences/Sofia2005.pdf},
year = {2005}
}

@article{katsaggelos2005advances,
abstract = {Multimedia applications involving the transmission of video over communication networks are rapidly increasing in popularity. Such applications can greatly benefit from adapting video coding parameters to network conditions as well as adapting network parameters to better support the application requirements. These two dimensions can both be viewed as allocating source and network resources to improve video quality. In this paper, we highlight recent advances in optimal resource allocation for real-time video communications over unreliable and resource constrained communication channels. More specifically, we focus on point-to-point coding and delivery schemes in which the sequences are encoded on the fly. We present a high-level framework for resource-distortion optimization. The framework can be used for jointly considering factors across network layers, including source coding, channel resource allocation, and error concealment. For example, resources can take the form of transmission energy in a wireless channel, and transmission cost in a DiffServ-based Internet channel. This framework can be used to optimally trade off resource consumption with end-to-end video quality in packet-based video transmission. After giving an overview of this framework, we review recent work in two areas - energy efficient wireless video transmission and resource allocation for Internet-based applications. {\textcopyright} 2005 IEEE.},
author = {Katsaggelos, A.K. and Eisenberg, Yiftach and Zhai, Fan and Berry, Randall and Pappas, T.N.},
doi = {10.1109/JPROC.2004.839621},
issn = {0018-9219},
journal = {Proceedings of the IEEE},
keywords = {Cross-layer design,Distortion estimation,Energy efficient,Error resilience,Internet video,Wireless video},
month = {jan},
number = {1},
pages = {135--147},
publisher = {IEEE},
title = {{Advances in Efficient Resource Allocation for Packet-Based Real-Time Video Transmission}},
url = {http://ieeexplore.ieee.org/document/1369704/},
volume = {93},
year = {2005}
}

Multimedia applications involving the transmission of video over communication networks are rapidly increasing in popularity. Such applications can greatly benefit from adapting video coding parameters to network conditions as well as adapting network parameters to better support the application requirements. These two dimensions can both be viewed as allocating source and network resources to improve video quality. In this paper, we highlight recent advances in optimal resource allocation for real-time video communications over unreliable and resource constrained communication channels. More specifically, we focus on point-to-point coding and delivery schemes in which the sequences are encoded on the fly. We present a high-level framework for resource-distortion optimization. The framework can be used for jointly considering factors across network layers, including source coding, channel resource allocation, and error concealment. For example, resources can take the form of transmission energy in a wireless channel, and transmission cost in a DiffServ-based Internet channel. This framework can be used to optimally trade off resource consumption with end-to-end video quality in packet-based video transmission. After giving an overview of this framework, we review recent work in two areas - energy efficient wireless video transmission and resource allocation for Internet-based applications. © 2005 IEEE.

@article{Aggelos2005,
abstract = {Transmitting video over wireless channels from mobile devices has gained increased popularity in a wide range of applications. A major obstacle in these types of applications is the limited energy supply in mobile device batteries. For this reason, efficiently utilizing energy is a critical issue in designing wireless video communication systems. This article highlights recent advances in joint source coding and optimal energy allocation. We present a general framework that takes into account multiple factors, including source coding, channel resource allocation, and error concealment, for the design of energy-efficient wireless video communication systems. This framework can take various forms and be applied to achieve the optimal trade-off between energy consumption and video delivery quality during wireless video transmission. {\textcopyright} 2005 IEEE.},
author = {Katsaggelos, A.K. and {Fan Zhai} and Eisenberg, Yiftach and Berry, Randall},
doi = {10.1109/MWC.2005.1497855},
issn = {1536-1284},
journal = {IEEE Wireless Communications},
month = {aug},
number = {4},
pages = {24--30},
title = {{Energy-efficient wireless video coding and delivery}},
url = {http://ieeexplore.ieee.org/document/1497855/},
volume = {12},
year = {2005}
}

Transmitting video over wireless channels from mobile devices has gained increased popularity in a wide range of applications. A major obstacle in these types of applications is the limited energy supply in mobile device batteries. For this reason, efficiently utilizing energy is a critical issue in designing wireless video communication systems. This article highlights recent advances in joint source coding and optimal energy allocation. We present a general framework that takes into account multiple factors, including source coding, channel resource allocation, and error concealment, for the design of energy-efficient wireless video communication systems. This framework can take various forms and be applied to achieve the optimal trade-off between energy consumption and video delivery quality during wireless video transmission. © 2005 IEEE.

@article{li2005rate,
abstract = {The need for video summarization originates primarily from a viewing time constraint. A shorter version of the original video sequence is desirable in a number of applications. Clearly, a shorter version is also necessary in applications where storage, communication bandwidth, and/or power are limited. The summarization process inevitably introduces distortion. The amount of summarization distortion is related to its "conciseness," or the number of frames available in the summary. If there are m frames in the original sequence and n frames in the summary, we define the summarization rate as m/n, to characterize this "conciseness." We also develop a new summarization distortion metric and formulate the summarization problem as a rate-distortion optimization problem. Optimal algorithms based on dynamic programming are presented and compared experimentally with heuristic algorithms. Practical constraints, like the maximum number of frames that can be skipped, are also considered in the formulation and solution of the problem. {\textcopyright} 2005 IEEE.},
author = {Li, Zhu and Schuster, G.M. Guido M. and Katsaggelos, Aggelos K. A.K. and Gandhi, Bhavan and {Zhu Li} and Schuster, G.M. Guido M. and Katsaggelos, Aggelos K. A.K. and Gandhi, Bhavan},
doi = {10.1109/TIP.2005.854477},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Dynamic programming (DP),Rate-distortion optimization,Video analysis,Video summarization},
month = {oct},
number = {10},
pages = {1550--1560},
pmid = {16238060},
publisher = {IEEE},
title = {{Rate-distortion optimal video summary generation}},
url = {http://ieeexplore.ieee.org/document/1510689/},
volume = {14},
year = {2005}
}

The need for video summarization originates primarily from a viewing time constraint. A shorter version of the original video sequence is desirable in a number of applications. Clearly, a shorter version is also necessary in applications where storage, communication bandwidth, and/or power are limited. The summarization process inevitably introduces distortion. The amount of summarization distortion is related to its "conciseness," or the number of frames available in the summary. If there are m frames in the original sequence and n frames in the summary, we define the summarization rate as m/n, to characterize this "conciseness." We also develop a new summarization distortion metric and formulate the summarization problem as a rate-distortion optimization problem. Optimal algorithms based on dynamic programming are presented and compared experimentally with heuristic algorithms. Practical constraints, like the maximum number of frames that can be skipped, are also considered in the formulation and solution of the problem. © 2005 IEEE.

@article{segall2004bayesian,
abstract = {Super-resolution algorithms recover high-frequency information from a sequence of low-resolution observations. In this paper, we consider the impact of video compression on the super-resolution task. Hybrid motion-compensation and transform coding schemes are the focus, as these methods provide observations of the underlying displacement values as well as a variable noise process. We utilize the Bayesian framework to incorporate this information and fuse the super-resolution and post-processing problems. A tractable solution is defined, and relationships between algorithm parameters and information in the compressed bitstream are established. The association between resolution recovery and compression ratio is also explored. Simulations illustrate the performance of the procedure with both synthetic and nonsynthetic sequences. {\textcopyright} 2004 IEEE.},
author = {Segall, C.A. and Katsaggelos, A.K. and Molina, Rafael and Mateos, Javier},
doi = {10.1109/TIP.2004.827230},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {jul},
number = {7},
pages = {898--911},
pmid = {15648857},
publisher = {IEEE},
title = {{Bayesian Resolution Enhancement of Compressed Video}},
url = {http://ieeexplore.ieee.org/document/1303643/},
volume = {13},
year = {2004}
}

Super-resolution algorithms recover high-frequency information from a sequence of low-resolution observations. In this paper, we consider the impact of video compression on the super-resolution task. Hybrid motion-compensation and transform coding schemes are the focus, as these methods provide observations of the underlying displacement values as well as a variable noise process. We utilize the Bayesian framework to incorporate this information and fuse the super-resolution and post-processing problems. A tractable solution is defined, and relationships between algorithm parameters and information in the compressed bitstream are established. The association between resolution recovery and compression ratio is also explored. Simulations illustrate the performance of the procedure with both synthetic and nonsynthetic sequences. © 2004 IEEE.

@article{Park2004,
author = {Park, S.C. and Kang, M.G. and Segall, C.A. and Katsaggelos, A.K.},
doi = {10.1109/TIP.2003.819233},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {apr},
number = {4},
pages = {573--585},
title = {{Spatially Adaptive High-Resolution Image Reconstruction of DCT-Based Compressed Images}},
url = {http://ieeexplore.ieee.org/document/1284393/},
volume = {13},
year = {2004}
}

@article{Sotirios2004a,
author = {Tsaftaris, S.A. and Katsaggelos, A.K. and Pappas, T.N. and Papoutsakis, E.T.},
doi = {10.1109/MSP.2004.1328093},
issn = {1053-5888},
journal = {IEEE Signal Processing Magazine},
month = {sep},
number = {5},
pages = {100--106},
title = {{Life sciences - DNA computing from a signal processing viewpoint}},
url = {http://ieeexplore.ieee.org/document/1328093/},
volume = {21},
year = {2004}
}

@article{aleksic2004speech,
abstract = {There is a strong correlation between the building blocks of speech (phonemes) and the building blocks of visual speech (visimes). In this paper, this correlation is exploited and an approach is proposed for synthesizing the visual representation of speech from a narrow-band acoustic speech signal. The visual speech is represented in terms of the facial animation parameters (FAPs), supported by the MPEG-4 standard. The main contribution of this paper is the development of a correlation hidden Markov model (CHMM) system, which integrates independently trained acoustic HMM (AHMM) and visual HMM (VHMM) systems, in order to realize speech-to-video synthesis. The proposed CHMM system allows for different model topologies for acoustic and visual HMMs. It performs late integration and reduces the amount of required training data compared to early integration modeling techniques. Temporal accuracy experiments, comparison of the synthesized FAPs to the original FAPs, and audio-visual automatic speech recognition (AV-ASR) experiments utilizing the synthesized visual speech were performed in order to objectively measure the performance of the system. The objective experiments demonstrated that the proposed approach reduces time alignment errors by 40.5% compared to the conventional temporal scaling method, that the synthesized FAP sequences are very similar to the original FAP sequences, and that synthesized FAP sequences contain visual speechreading information that can improve AV-ASR performance.},
author = {Aleksic, P.S. and Katsaggelos, A.K.},
doi = {10.1109/TCSVT.2004.826760},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
keywords = {Audio-visual speech recognition,Correlation hidden Markov models (CHMMs),Facial animation parameters (FAPs),Speech-to-video synthesis},
month = {may},
number = {5},
pages = {682--692},
publisher = {IEEE},
title = {{Speech-To-Video Synthesis Using MPEG-4 Compliant Visual Features}},
url = {http://ieeexplore.ieee.org/document/1294959/},
volume = {14},
year = {2004}
}

There is a strong correlation between the building blocks of speech (phonemes) and the building blocks of visual speech (visimes). In this paper, this correlation is exploited and an approach is proposed for synthesizing the visual representation of speech from a narrow-band acoustic speech signal. The visual speech is represented in terms of the facial animation parameters (FAPs), supported by the MPEG-4 standard. The main contribution of this paper is the development of a correlation hidden Markov model (CHMM) system, which integrates independently trained acoustic HMM (AHMM) and visual HMM (VHMM) systems, in order to realize speech-to-video synthesis. The proposed CHMM system allows for different model topologies for acoustic and visual HMMs. It performs late integration and reduces the amount of required training data compared to early integration modeling techniques. Temporal accuracy experiments, comparison of the synthesized FAPs to the original FAPs, and audio-visual automatic speech recognition (AV-ASR) experiments utilizing the synthesized visual speech were performed in order to objectively measure the performance of the system. The objective experiments demonstrated that the proposed approach reduces time alignment errors by 40.5% compared to the conventional temporal scaling method, that the synthesized FAP sequences are very similar to the original FAP sequences, and that synthesized FAP sequences contain visual speechreading information that can improve AV-ASR performance.

@article{Ebroul2004,
author = {Izquierdo, Ebroul and Katsaggelos, A.K. and Strintzis, M.G.},
doi = {10.1109/TCSVT.2004.828719},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
month = {may},
number = {5},
pages = {569--571},
title = {{Introduction to the Special Issue on Audio and Video Analysis for Multimedia Interactive Services}},
url = {http://ieeexplore.ieee.org/document/1294949/},
volume = {14},
year = {2004}
}

@article{schuster2004shape,
abstract = {The introduction of video objects (VOs) is one of the innovations of MPEG-4. The $\alpha$-plane of a VO defines its shape at a given instance in time and hence determines the boundary of its texture. In packet-based networks, shape, motion, and texture are subject to loss. While there has been considerable attention paid to the concealment of texture and motion errors, little has been done in the field of shape error concealment. In this paper, we propose a post-processing shape error-concealment technique that uses geometric boundary information of the received $\alpha$-plane. Second-order Hermite splines are used to model the received boundary in the neighboring blocks, while third order Hermite splines are used to model the missing boundary. The velocities of these splines are matched at the boundary point closest to the missing block. There exists the possibility of multiple concealing splines per group of lost boundary parts. Therefore, we draw every concealment spline combination that does not self-intersect and keep all possible results until the end. At the end, we select the concealment solution that results in one closed boundary. Experimental results demonstrating the performance of the proposed method and comparisons with prior proposed methods are presented.},
author = {Schuster, G.M. and Li, Xiaohuan and Katsaggelos, A.K.},
doi = {10.1109/TIP.2004.827226},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Error concealment,Hermite spline,MPEG-4,Shape coding},
month = {jun},
number = {6},
pages = {808--820},
pmid = {15648871},
publisher = {IEEE},
title = {{Shape Error Concealment Using Hermite Splines}},
url = {http://ieeexplore.ieee.org/document/1298837/},
volume = {13},
year = {2004}
}

The introduction of video objects (VOs) is one of the innovations of MPEG-4. The $α$-plane of a VO defines its shape at a given instance in time and hence determines the boundary of its texture. In packet-based networks, shape, motion, and texture are subject to loss. While there has been considerable attention paid to the concealment of texture and motion errors, little has been done in the field of shape error concealment. In this paper, we propose a post-processing shape error-concealment technique that uses geometric boundary information of the received $α$-plane. Second-order Hermite splines are used to model the received boundary in the neighboring blocks, while third order Hermite splines are used to model the missing boundary. The velocities of these splines are matched at the boundary point closest to the missing block. There exists the possibility of multiple concealing splines per group of lost boundary parts. Therefore, we draw every concealment spline combination that does not self-intersect and keep all possible results until the end. At the end, we select the concealment solution that results in one closed boundary. Experimental results demonstrating the performance of the proposed method and comparisons with prior proposed methods are presented.

@article{tsaftaris2004can,
author = {Tsaftaris, S.A. and Katsaggelos, A.K. and Pappas, T.N. and Papoutsakis, E.T.},
doi = {10.1109/MSP.2004.1359142},
issn = {1053-5888},
journal = {IEEE Signal Processing Magazine},
month = {nov},
number = {6},
pages = {57--61},
publisher = {IEEE},
title = {{How can DNA computing be applied to digital signal processing?}},
url = {http://ieeexplore.ieee.org/document/1359142/},
volume = {21},
year = {2004}
}

@article{alvarez2004high,
abstract = {In this article, we address the problem of obtaining a high-resolution (HR) image from a compressed low-resolution (LR) video sequence. Motion information plays a critical role in solving this problem, and we determine which pixels in the sequence provide useful information for calculating the high-resolution image. The bit stream of hybrid motion compensated video compression methods includes low-resolution motion-compensated images; we therefore also study which pixels in these images should be used to increase the quality of the reconstructed image. Once the useful (observable) pixels in the low-resolution and motion-compensated sequences have been detected, we modify the acquisition model to only account for these observations. The proposed approach is tested on real compressed video sequences and the improved performance is re ported. {\textcopyright} 2004 Wiley Periodicals, Inc.},
author = {Alvarez, L. D. and Mateos, J. and Molina, R. and Katsaggelos, A. K.},
doi = {10.1002/ima.20008},
issn = {0899-9457},
journal = {International Journal of Imaging Systems and Technology},
keywords = {Compressed low-resolution video,High-resolution image,Motion estimation,Observability map,Observable pixel},
number = {2},
pages = {58--66},
publisher = {Wiley Subscription Services, Inc., A Wiley Company Hoboken},
title = {{High-resolution images from compressed low-resolution video: Motion estimation and observable pixels}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/ima.20008},
volume = {14},
year = {2004}
}

In this article, we address the problem of obtaining a high-resolution (HR) image from a compressed low-resolution (LR) video sequence. Motion information plays a critical role in solving this problem, and we determine which pixels in the sequence provide useful information for calculating the high-resolution image. The bit stream of hybrid motion compensated video compression methods includes low-resolution motion-compensated images; we therefore also study which pixels in these images should be used to increase the quality of the reconstructed image. Once the useful (observable) pixels in the low-resolution and motion-compensated sequences have been detected, we modify the acquisition model to only account for these observations. The proposed approach is tested on real compressed video sequences and the improved performance is re ported. © 2004 Wiley Periodicals, Inc.

@article{Antonio2004,
author = {L{\'{o}}pez, Antonio and Molina, Rafael and Katsaggelos, Aggelos K. and Rodriguez, Antonio and L{\'{o}}pez, Jos{\'{e}} M. and Llamas, Jos{\'{e}} M.},
doi = {10.1002/ima.20003},
issn = {08999457},
journal = {International Journal of Imaging Systems and Technology},
number = {1},
pages = {21--27},
title = {{Parameter estimation in Bayesian reconstruction of SPECT images: An aid in nuclear medicine diagnosis}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/ima.20003},
volume = {14},
year = {2004}
}

@article{Tom2004,
abstract = {This paper presents a method to find the operational rate-distortion optimal solution for an overcomplete signal decomposition. The idea of using overcomplete dictionaries, or frames, is to get a sparse representation of the signal. Traditionally, sub-optimal algorithms, such as matching pursuit (MP), are used for this purpose. When using frames in a lossy compression scheme, the major issue is to find the best possible rate-distortion (RD) tradeoff. Given the frame and the variable length code (VLC) table embedded in the entropy coder, the solution to the problem of establishing the best RD tradeoff is highly complex. The proposed approach reduces this complexity significantly by structuring the solution approach such that the dependent quantizer allocation problem reduces to an independent one. In addition, the use of a solution tree further reduces the complexity. It is important to note that this large reduction in complexity is achieved without sacrificing optimality. The optimal rate-distortion solution depends on the selection of the frame and the VLC table embedded in the entropy coder. Thus, frame design and VLC optimization is part of this work. We experimentally demonstrate that the new approach outperforms rate-distortion optimized (RDO) matching pursuit, previously proposed by Gharavi-Alkhansari.},
author = {Ryen, Tom and Schuster, G.M. and Katsaggelos, A.K.},
doi = {10.1109/TSP.2004.826184},
issn = {1053-587X},
journal = {IEEE Transactions on Signal Processing},
keywords = {Depth-first search,Frame design,Overcomplete dictionary,QR-decomposition,Rate-distortion optimization},
month = {may},
number = {5},
pages = {1352--1363},
title = {{A Rate-Distortion Optimal Alternative to Matching Pursuit}},
url = {https://ieeexplore.ieee.org/iel5/78/28675/01284833.pdf http://ieeexplore.ieee.org/document/1284833/},
volume = {52},
year = {2004}
}

This paper presents a method to find the operational rate-distortion optimal solution for an overcomplete signal decomposition. The idea of using overcomplete dictionaries, or frames, is to get a sparse representation of the signal. Traditionally, sub-optimal algorithms, such as matching pursuit (MP), are used for this purpose. When using frames in a lossy compression scheme, the major issue is to find the best possible rate-distortion (RD) tradeoff. Given the frame and the variable length code (VLC) table embedded in the entropy coder, the solution to the problem of establishing the best RD tradeoff is highly complex. The proposed approach reduces this complexity significantly by structuring the solution approach such that the dependent quantizer allocation problem reduces to an independent one. In addition, the use of a solution tree further reduces the complexity. It is important to note that this large reduction in complexity is achieved without sacrificing optimality. The optimal rate-distortion solution depends on the selection of the frame and the VLC table embedded in the entropy coder. Thus, frame design and VLC optimization is part of this work. We experimentally demonstrate that the new approach outperforms rate-distortion optimized (RDO) matching pursuit, previously proposed by Gharavi-Alkhansari.

@article{Lisimachos2004,
abstract = {A major problem in object-oriented video coding and MPEG-4 is the encoding of object boundaries. Traditionally this problem is treated separately from the texture encoding problem. In this paper, we present a vertex-based shape coding method which is optimal in the operational rate-distortion sense and takes into account the texture information of the video frames. This is accomplished by utilizing a variable-width tolerance band whose width is a function of the texture profile. As an example, this width is inversely proportional to the magnitude of the image gradient. Thus, in areas where the confidence in the estimation of the boundary is low and/or coding errors in the boundary will not affect the application (e.g., object-oriented coding and MPEG-4) significantly, a larger boundary approximation error is allowed. We present experimental results which demonstrate the effectiveness of the proposed algorithm.},
author = {Kondi, L.P. and Melnikov, Gerry and Katsaggelos, A.K.},
doi = {10.1109/TCSVT.2004.825569},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
keywords = {Boundary coding,MPEG-4,Operational rate-distortion,Operational rate-distortion (ORD) theory,Shape coding,Shape-adaptive DCT},
month = {apr},
number = {4},
pages = {528--533},
publisher = {IEEE},
title = {{Joint Optimal Object Shape Estimation and Encoding}},
url = {http://ieeexplore.ieee.org/document/1281827/},
volume = {14},
year = {2004}
}

A major problem in object-oriented video coding and MPEG-4 is the encoding of object boundaries. Traditionally this problem is treated separately from the texture encoding problem. In this paper, we present a vertex-based shape coding method which is optimal in the operational rate-distortion sense and takes into account the texture information of the video frames. This is accomplished by utilizing a variable-width tolerance band whose width is a function of the texture profile. As an example, this width is inversely proportional to the magnitude of the image gradient. Thus, in areas where the confidence in the estimation of the boundary is low and/or coding errors in the boundary will not affect the application (e.g., object-oriented coding and MPEG-4) significantly, a larger boundary approximation error is allowed. We present experimental results which demonstrate the effectiveness of the proposed algorithm.

@article{nakagaki2003vq,
abstract = {In this paper, learning-based algorithms for image restoration and blind image restoration are proposed. Such algorithms deviate from the traditional approaches in this area, by utilizing priors that are learned from similar images. Original images and their degraded versions by the known degradation operator (restoration problem) are utilized for designing the VQ codebooks. The codevectors are designed using the blurred images. For each such vector, the high frequency information obtained from the original images is also available. During restoration, the high frequency information of a given degraded image is estimated from its low frequency information based on the codebooks. For the blind restoration problem, a number of codebooks are designed corresponding to various versions of the blurring function. Given a noisy and blurred image, one of the codebooks is chosen based on a similarity measure, therefore providing the identification of the blur. To make the restoration process computationally efficient, the Principal Component Analysis (PCA) and VQ-Nearest Neighborhood approaches are utilized. Simulation results are presented to demonstrate the effectiveness of the proposed algorithms.},
author = {Nakagaki, Ryo and Katsaggelos, A.K.},
doi = {10.1109/TIP.2003.816007},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Blur identification,Compression,Image restoration,Vector quantization},
month = {sep},
number = {9},
pages = {1044--1053},
publisher = {IEEE},
title = {{A VQ-based blind image restoration algorithm}},
url = {http://ieeexplore.ieee.org/document/1221758/},
volume = {12},
year = {2003}
}

In this paper, learning-based algorithms for image restoration and blind image restoration are proposed. Such algorithms deviate from the traditional approaches in this area, by utilizing priors that are learned from similar images. Original images and their degraded versions by the known degradation operator (restoration problem) are utilized for designing the VQ codebooks. The codevectors are designed using the blurred images. For each such vector, the high frequency information obtained from the original images is also available. During restoration, the high frequency information of a given degraded image is estimated from its low frequency information based on the codebooks. For the blind restoration problem, a number of codebooks are designed corresponding to various versions of the blurring function. Given a noisy and blurred image, one of the codebooks is chosen based on a similarity measure, therefore providing the identification of the blur. To make the restoration process computationally efficient, the Principal Component Analysis (PCA) and VQ-Nearest Neighborhood approaches are utilized. Simulation results are presented to demonstrate the effectiveness of the proposed algorithms.

@article{Luna2003,
author = {Luna, C.E. and Eisenberg, Y. and Berry, R. and Pappas, T.N. and Katsaggelos, A.K.},
doi = {10.1109/JSAC.2003.815394},
issn = {0733-8716},
journal = {IEEE Journal on Selected Areas in Communications},
month = {dec},
number = {10},
pages = {1710--1720},
title = {{Joint source coding and data rate adaptation for energy efficient wireless video streaming}},
url = {http://ieeexplore.ieee.org/document/1254586/},
volume = {21},
year = {2003}
}

@article{wang2003efficient,
abstract = {In this paper, we present a new shape-coding approach, which decouples the shape information into two independent signal data sets; the skeleton and the boundary distance from the skeleton. The major benefit of this approach is that it allows for a more flexible tradeoff between approximation error and bit budget. Curves of arbitrary order can be utilized for approximating both the skeleton and distance signals. For a given bit budget for a video frame, we solve the problem of choosing the number and location of the control points for all skeleton and distance signals of all boundaries within a frame, so that the overall distortion is minimized. An operational rate-distortion (ORD) optimal approach using Lagrangian relaxation and a four-dimensional Direct Acyclic Graph (DAG) shortest path algorithm is developed for solving the problem. To reduce the computational complexity from O(N5) to O(N3), where N is the number of admissible control points for a skeleton, a suboptimal greedy-trellis search algorithm is proposed and compared with the optimal algorithm. In addition, an even more efficient algorithm with computational complexity O(N2) that finds an ORD optimal solution using a relaxed distortion criterion is also proposed and compared with the optimal solution. Experimental results demonstrate that our proposed approaches outperform existing ORD optimal approaches, which do not follow the same decomposition of the source data.},
author = {{Haohong Wang} and Schuster, G.M. Guido M. and Katsaggelos, Aggelos K. A.K. and Pappas, T.N. Thrasyvoulos N. and Wang, Haohong and Schuster, G.M. Guido M. and Katsaggelos, Aggelos K. A.K. and Pappas, T.N. Thrasyvoulos N.},
doi = {10.1109/TIP.2003.816570},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Boundary coding,Object-based video compression,Rate-distortion optimization,Shape coding,Skeleton decomposition,Skeletonization},
month = {oct},
number = {10},
pages = {1181--1193},
publisher = {IEEE},
title = {{An Efficient Rate-Distortion Optimal Shape Coding Approach Utilizing a Skeleton-Based Decomposition}},
url = {http://ieeexplore.ieee.org/document/1233561/},
volume = {12},
year = {2003}
}

In this paper, we present a new shape-coding approach, which decouples the shape information into two independent signal data sets; the skeleton and the boundary distance from the skeleton. The major benefit of this approach is that it allows for a more flexible tradeoff between approximation error and bit budget. Curves of arbitrary order can be utilized for approximating both the skeleton and distance signals. For a given bit budget for a video frame, we solve the problem of choosing the number and location of the control points for all skeleton and distance signals of all boundaries within a frame, so that the overall distortion is minimized. An operational rate-distortion (ORD) optimal approach using Lagrangian relaxation and a four-dimensional Direct Acyclic Graph (DAG) shortest path algorithm is developed for solving the problem. To reduce the computational complexity from O(N5) to O(N3), where N is the number of admissible control points for a skeleton, a suboptimal greedy-trellis search algorithm is proposed and compared with the optimal algorithm. In addition, an even more efficient algorithm with computational complexity O(N2) that finds an ORD optimal solution using a relaxed distortion criterion is also proposed and compared with the optimal solution. Experimental results demonstrate that our proposed approaches outperform existing ORD optimal approaches, which do not follow the same decomposition of the source data.

@article{luna2003maximizing,
abstract = {In this paper, we study some of the design tradeoffs of video streaming systems in networks with QoS guarantees. We approach this problem by using a utility function to quantify the benefit a user derives from the quality of the received video sequence. We also consider the cost to the network user for streaming the video sequence. We have formulated this utility maximization problem as a joint constrained optimization problem where we maximize the difference between the utility and the network cost, subject to the constraint that the decoder buffer does not underflow. In this manner, we can find the optimal tradeoff between video quality and network cost. We present a deterministic dynamic programming approach for both the constant bit rate and renegotiated constant bit rate service classes. Experimental results demonstrate the benefits and the performance of the proposed approach.},
author = {Luna, C.E. and Kondi, L.P. and Katsaggelos, A.K.},
doi = {10.1109/TCSVT.2002.808439},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
keywords = {Quality of service (QoS),Renegotiated constant bit rate (RCBR),Renegotiated services,User utility,Video streaming},
month = {feb},
number = {2},
pages = {141--148},
publisher = {IEEE},
title = {{Maximizing user utility in video streaming applications}},
url = {http://ieeexplore.ieee.org/document/1186530/},
volume = {13},
year = {2003}
}

In this paper, we study some of the design tradeoffs of video streaming systems in networks with QoS guarantees. We approach this problem by using a utility function to quantify the benefit a user derives from the quality of the received video sequence. We also consider the cost to the network user for streaming the video sequence. We have formulated this utility maximization problem as a joint constrained optimization problem where we maximize the difference between the utility and the network cost, subject to the constraint that the decoder buffer does not underflow. In this manner, we can find the optimal tradeoff between video quality and network cost. We present a deterministic dynamic programming approach for both the constant bit rate and renegotiated constant bit rate service classes. Experimental results demonstrate the benefits and the performance of the proposed approach.

@article{molina2003parameter,
abstract = {In this paper, we consider the estimation of the unknown parameters for the problem of reconstructing a high-resolution image from multiple undersampled, shifted, degraded frames with subpixel displacement errors. We derive mathematical expressions for the iterative calculation of the maximum likelihood estimate of the unknown parameters given the low resolution observed images. These iterative procedures require the manipulation of block-semi circulant (BSC) matrices, that is block matrices with circulant blocks. We show how these BSC matrices can be easily manipulated in order to calculate the unknown parameters. Finally the proposed method is tested on real and synthetic images.},
author = {Molina, Rafael and Vega, Miguel and Abad, Javier and Katsaggelos, A.K.},
doi = {10.1109/TIP.2003.818117},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian methods,High-resolution image reconstruction,Parameter estimation},
month = {dec},
number = {12},
pages = {1655--1667},
publisher = {IEEE},
title = {{Parameter estimation in bayesian high-resolution image reconstruction with multisensors}},
url = {http://ieeexplore.ieee.org/document/1257401/},
volume = {12},
year = {2003}
}

In this paper, we consider the estimation of the unknown parameters for the problem of reconstructing a high-resolution image from multiple undersampled, shifted, degraded frames with subpixel displacement errors. We derive mathematical expressions for the iterative calculation of the maximum likelihood estimate of the unknown parameters given the low resolution observed images. These iterative procedures require the manipulation of block-semi circulant (BSC) matrices, that is block matrices with circulant blocks. We show how these BSC matrices can be easily manipulated in order to calculate the unknown parameters. Finally the proposed method is tested on real and synthetic images.

@article{Molina2003,
author = {Molina, R. and Mateos, J. and Katsaggelos, A.K. and Vega, M.},
doi = {10.1109/TIP.2003.818015},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {dec},
number = {12},
pages = {1642--1654},
title = {{Bayesian multichannel image restoration using compound gauss-markov random fields}},
url = {http://ieeexplore.ieee.org/document/1257400/},
volume = {12},
year = {2003}
}

@article{Petar2003a,
author = {Aleksic, Petar S and Katsaggelos, Aggelos K},
journal = {Works. Multimedia User Authentication, Santa Barbara, CA},
title = {{An audio-visual person identification and verification system using FAPs as visual features}},
url = {https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=fd43812e63659d50b2f96c6f5f361a1dc64f61fe#page=80},
year = {2003}
}

@article{Kaaren2003,
abstract = {Spatially adaptive intensity bounds on the image estimate are shown to be an effective means of regularising the ill-posed image restoration problem. For blind restoration, the local intensity constraints also help to further define the solution, thereby reducing the number of multiple solutions and local minima. The bounds are defined in terms of the local statistics of the image estimate and a control parameter which determines the scale of the bounds. Guidelines for choosing this parameter are developed in the context of classical (nonblind) image restoration. The intensity bounds are applied by means of the gradient projection method, and conditions for convergence are derived when the bounds are refined using the current image estimate. Based on this method, a new alternating constrained minimisation approach is proposed for blind image restoration. On the basis of the experimental results provided, it is found that local intensity bounds offer a simple, flexible method of constraining both the nonblind and blind restoration problems.},
author = {May, Kaaren L. and Stathaki, Tania and Katsaggelos, Aggelos K.},
doi = {10.1155/S1110865703308066},
issn = {1687-6180},
journal = {EURASIP Journal on Advances in Signal Processing},
keywords = {Blind image restoration,Blur identification,Image resolution,Set-theoretic estimation},
month = {dec},
number = {12},
pages = {250181},
title = {{Spatially Adaptive Intensity Bounds for Image Restoration}},
url = {https://asp-eurasipjournals.springeropen.com/articles/10.1155/S1110865703308066},
volume = {2003},
year = {2003}
}

Spatially adaptive intensity bounds on the image estimate are shown to be an effective means of regularising the ill-posed image restoration problem. For blind restoration, the local intensity constraints also help to further define the solution, thereby reducing the number of multiple solutions and local minima. The bounds are defined in terms of the local statistics of the image estimate and a control parameter which determines the scale of the bounds. Guidelines for choosing this parameter are developed in the context of classical (nonblind) image restoration. The intensity bounds are applied by means of the gradient projection method, and conditions for convergence are derived when the bounds are refined using the current image estimate. Based on this method, a new alternating constrained minimisation approach is proposed for blind image restoration. On the basis of the experimental results provided, it is found that local intensity bounds offer a simple, flexible method of constraining both the nonblind and blind restoration problems.

@article{segall2003high,
author = {Segall, C.A. and Molina, Rafael and Katsaggelos, A.K.},
doi = {10.1109/MSP.2003.1203208},
issn = {1053-5888},
journal = {IEEE Signal Processing Magazine},
month = {may},
number = {3},
pages = {37--48},
publisher = {IEEE},
title = {{High-resolution images from low-resolution compressed video}},
url = {http://ieeexplore.ieee.org/document/1203208/},
volume = {20},
year = {2003}
}

@article{SedaOgrenci2003,
abstract = {Programmable logic is emerging as an attractive solution for many digital signal processing applications. In this work, we have investigated issues arising due to the resource constraints of FPGA-based systems. Using an iterative image restoration algorithm as an example we have shown how to manipulate the original algorithm to suit it to an FPGA implementation. Consequences of such manipulations have been estimated, such as loss of quality in the output image. We also present performance results from an actual implementation on a Xilinx FPGA. Our experiments demonstrate that, for different criteria, such as result quality or speed, the best implementation is different as well.},
author = {Memik, S.O. and Katsaggelos, A.K. and Sarrafzadeh, Majid},
doi = {10.1109/TC.2003.1183952},
issn = {0018-9340},
journal = {IEEE Transactions on Computers},
keywords = {FPGA,Image restoration,Image segmentation},
month = {mar},
number = {3},
pages = {390--399},
title = {{Analysis and FPGA implementation of image restoration under resource constraints}},
url = {http://ieeexplore.ieee.org/document/1183952/},
volume = {52},
year = {2003}
}

Programmable logic is emerging as an attractive solution for many digital signal processing applications. In this work, we have investigated issues arising due to the resource constraints of FPGA-based systems. Using an iterative image restoration algorithm as an example we have shown how to manipulate the original algorithm to suit it to an FPGA implementation. Consequences of such manipulations have been estimated, such as loss of quality in the output image. We also present performance results from an actual implementation on a Xilinx FPGA. Our experiments demonstrate that, for different criteria, such as result quality or speed, the best implementation is different as well.

@article{Eisenberg2002,
author = {Eisenberg, Y. and Luna, C.E. and Pappas, T.N. and Berry, R. and Katsaggelos, A.K.},
doi = {10.1109/TCSVT.2002.800309},
issn = {1051-8215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
month = {jun},
number = {6},
pages = {411--424},
title = {{Joint source coding and transmission power management for energy efficient wireless video communications}},
url = {http://ieeexplore.ieee.org/document/1013849/},
volume = {12},
year = {2002}
}

@article{hong2002iterative,
author = {Hong, Min-Cheol and Stathaki, Tania and Katsaggelos, Aggelos K.},
doi = {10.1117/1.1503072},
issn = {0091-3286},
journal = {Optical Engineering},
month = {oct},
number = {10},
pages = {2515},
publisher = {Society of Photo-Optical Instrumentation Engineers},
title = {{Iterative regularized least-mean mixed-norm image restoration}},
url = {http://opticalengineering.spiedigitallibrary.org/article.aspx?doi=10.1117/1.1503072},
volume = {41},
year = {2002}
}

@article{Petar2002,
abstract = {We describe an audio-visual automatic continuous speech recognition system, which significantly improves speech recognition performance over a wide range of acoustic noise levels, as well as under clean audio conditions. The system utilizes facial animation parameters (FAPs) supported by the MPEG-4 standard for the visual representation of speech. We also describe a robust and automatic algorithm we have developed to extract FAPs from visual data, which does not require hand labeling or extensive training procedures. The principal component analysis (PCA) was performed on the FAPs in order to decrease the dimensionality of the visual feature vectors, and the derived projection weights were used as visual features in the audio-visual automatic speech recognition (ASR) experiments. Both single-stream and multistream hidden Markov models (HMMs) were used to model the ASR system, integrate audio and visual information, and perform a relatively large vocabulary (approximately 1000 words) speech recognition experiments. The experiments performed use clean audio data and audio data corrupted by stationary white Gaussian noise at various SNRs. The proposed system reduces the word error rate (WER) by 20% to 23% relatively to audio-only speech recognition WERs, at various SNRs (0-30 dB) with additive white Gaussian noise, and by 19% relatively to audio-only speech recognition WER under clean audio conditions.},
author = {Aleksic, Petar S. and Williams, Jay J. and Wu, Zhilin and Katsaggelos, Aggelos K.},
doi = {10.1155/S1110865702206162},
issn = {1687-6180},
journal = {EURASIP Journal on Advances in Signal Processing},
keywords = {Audio-visual speech recognition,Facial animation parameters,Snake},
month = {dec},
number = {11},
pages = {150948},
publisher = {Springer International Publishing},
title = {{Audio-Visual Speech Recognition Using MPEG-4 Compliant Visual Features}},
url = {https://asp-eurasipjournals.springeropen.com/articles/10.1155/S1110865702206162},
volume = {2002},
year = {2002}
}

We describe an audio-visual automatic continuous speech recognition system, which significantly improves speech recognition performance over a wide range of acoustic noise levels, as well as under clean audio conditions. The system utilizes facial animation parameters (FAPs) supported by the MPEG-4 standard for the visual representation of speech. We also describe a robust and automatic algorithm we have developed to extract FAPs from visual data, which does not require hand labeling or extensive training procedures. The principal component analysis (PCA) was performed on the FAPs in order to decrease the dimensionality of the visual feature vectors, and the derived projection weights were used as visual features in the audio-visual automatic speech recognition (ASR) experiments. Both single-stream and multistream hidden Markov models (HMMs) were used to model the ASR system, integrate audio and visual information, and perform a relatively large vocabulary (approximately 1000 words) speech recognition experiments. The experiments performed use clean audio data and audio data corrupted by stationary white Gaussian noise at various SNRs. The proposed system reduces the word error rate (WER) by 20% to 23% relatively to audio-only speech recognition WERs, at various SNRs (0-30 dB) with additive white Gaussian noise, and by 19% relatively to audio-only speech recognition WER under clean audio conditions.

@article{kondi2002joint,
abstract = {In this paper, we develop an approach toward joint source-channel coding for motion-compensated DCT-based scalable video coding and transmission. A framework for the optimal selection of the source and channel coding rates over all scalable layers is presented such that the overall distortion is minimized. The algorithm utilizes universal rate distortion characteristics which are obtained experimentally and show the sensitivity of the source encoder and decoder to channel errors. The proposed algorithm allocates the available bit rate between scalable layers and, within each layer, between source and channel coding. We present the results of this rate allocation algorithm for video transmission over a wireless channel using the H.263 Version 2 signal-to-noise ratio (SNR) scalable codec for source coding and rate-compatible punctured convolutional (RCPC) codes for channel coding. We discuss the performance of the algorithm with respect to the channel conditions, coding methodologies, layer rates, and number of layers.},
author = {Kondi, L.P. and Ishtiaq, Faisal and Katsaggelos, A.K.},
doi = {10.1109/TIP.2002.802507},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Joint source-channel coding,Scalable video,Wireless video},
month = {sep},
number = {9},
pages = {1043--1052},
publisher = {IEEE},
title = {{Joint source-channel coding for motion-compensated DCT-based SNR scalable video}},
url = {http://ieeexplore.ieee.org/document/1036052/},
volume = {11},
year = {2002}
}

In this paper, we develop an approach toward joint source-channel coding for motion-compensated DCT-based scalable video coding and transmission. A framework for the optimal selection of the source and channel coding rates over all scalable layers is presented such that the overall distortion is minimized. The algorithm utilizes universal rate distortion characteristics which are obtained experimentally and show the sensitivity of the source encoder and decoder to channel errors. The proposed algorithm allocates the available bit rate between scalable layers and, within each layer, between source and channel coding. We present the results of this rate allocation algorithm for video transmission over a wireless channel using the H.263 Version 2 signal-to-noise ratio (SNR) scalable codec for source coding and rate-compatible punctured convolutional (RCPC) codes for channel coding. We discuss the performance of the algorithm with respect to the channel conditions, coding methodologies, layer rates, and number of layers.

@article{segall2002super,
address = {Boston},
author = {Segall, C Andrew and Katsaggelos, Aggelos K and Molina, Rafael and Mateos, Javier},
doi = {10.1007/0-306-47004-7_9},
journal = {Super-Resolution Imaging},
pages = {211--242},
publisher = {Kluwer Academic Publishers},
title = {{Super-Resolution from Compressed Video}},
url = {http://link.springer.com/10.1007/0-306-47004-7_9},
year = {2002}
}

@article{Jay2002,
author = {Williams, J.J. and Katsaggelos, A.K.},
doi = {10.1109/TNN.2002.1021891},
issn = {1045-9227},
journal = {IEEE Transactions on Neural Networks},
month = {jul},
number = {4},
pages = {900--915},
title = {{An HMM-based speech-to-video synthesizer}},
url = {http://ieeexplore.ieee.org/document/1021891/},
volume = {13},
year = {2002}
}

@article{Antonio2002,
abstract = {We propose a new iterative method for Maximum a Posteriori (MAP) reconstruction of SPECT (Single Photon Emission Computed Tomography) images. The method uses Compound Gauss Markov Random Fields (CGMRF) as prior model and is stochastic for the line process and deterministic for the reconstruction. Synthetic and real images are used to compare the new method with existing ones.},
author = {L{\'{O}}PEZ, ANTONIO and MOLINA, RAFAEL and MATEOS, JAVIER and KATSAGGELOS, AGGELOS K.},
doi = {10.1142/S0218001402001708},
issn = {0218-0014},
journal = {International Journal of Pattern Recognition and Artificial Intelligence},
keywords = {Bayesian reconstruction,Compound Gauss Markov random fields,Deterministic image reconstruction,SPECT imaging,Simulated annealing},
month = {may},
number = {03},
pages = {317--330},
title = {{SPECT IMAGE RECONSTRUCTION USING COMPOUND PRIOR MODELS}},
url = {https://www.worldscientific.com/doi/abs/10.1142/S0218001402001708},
volume = {16},
year = {2002}
}

We propose a new iterative method for Maximum a Posteriori (MAP) reconstruction of SPECT (Single Photon Emission Computed Tomography) images. The method uses Compound Gauss Markov Random Fields (CGMRF) as prior model and is stochastic for the line process and deterministic for the reconstruction. Synthetic and real images are used to compare the new method with existing ones.

@article{Gerry2002,
abstract = {In this paper a hybrid fractal and Discrete Cosine transform (DCT) coder is developed. Drawing on the ability of DCT to remove inter-pixel redundancies and on the ability of fractal transforms to capitalize on long-range correlations within the image, the hybrid coder performs an operationally optimal, in the rate-distortion sense, bit allocation among coding parameters. An orthogonal basis framework is used within which an image segmentation and a hybrid block-based transform are selected jointly. The selection of coefficients in the DCT component of the overall block transform is made a part of the optimization procedure. A Lagrangian multiplier approach is used to optimize the hybrid transform parameters together with the segmentation. Differential encoding of the DC coefficient is employed, with the scanning path based on a 3rd-order Hilbert curve. Simulation results show a significant improvement in quality with respect to the JPEG standard, an approach based on optimization of DCT basis vectors, as well as, the purely fractal techniques.},
author = {Melnikov, Gerry and Katsaggelos, A.K.},
doi = {10.1109/TMM.2002.806531},
issn = {1520-9210},
journal = {IEEE Transactions on Multimedia},
month = {dec},
number = {4},
pages = {413--422},
title = {{A jointly optimal fractal/DCT compression scheme}},
url = {http://ieeexplore.ieee.org/document/1176940/},
volume = {4},
year = {2002}
}

In this paper a hybrid fractal and Discrete Cosine transform (DCT) coder is developed. Drawing on the ability of DCT to remove inter-pixel redundancies and on the ability of fractal transforms to capitalize on long-range correlations within the image, the hybrid coder performs an operationally optimal, in the rate-distortion sense, bit allocation among coding parameters. An orthogonal basis framework is used within which an image segmentation and a hybrid block-based transform are selected jointly. The selection of coefficients in the DCT component of the overall block transform is made a part of the optimization procedure. A Lagrangian multiplier approach is used to optimize the hybrid transform parameters together with the segmentation. Differential encoding of the DC coefficient is employed, with the scanning path based on a 3rd-order Hilbert curve. Simulation results show a significant improvement in quality with respect to the JPEG standard, an approach based on optimization of DCT basis vectors, as well as, the purely fractal techniques.

@article{galatsanos2002hyperparameter,
abstract = {This work is motivated by the observation that it is not possible to reliably estimate simultaneously all the necessary hyperparameters in an image restoration problem when the point-spread function is assumed to be the sum of a known deterministic and an unknown random component. To solve this problem we propose to use gamma hyperpriors for the unknown hyperparameters. Two iterative algorithms that simultaneously restore the image and estimate the hyperparameters are derived, based on the application of evidence analysis within the hierarchical Bayesian framework. Numerical experiments are presented that show the benefits of introducing hyperpriors for this problem. {\textcopyright} 2002 Society of Photo-Optical Instrumentation Engineers.},
author = {Galatsanos, Nikolas P. and Mesarovic, Vladimir Z. and Molina, Rafael and Katsaggelos, Aggelos K. and Mateos, Javier},
doi = {10.1117/1.1487850},
issn = {0091-3286},
journal = {Optical Engineering},
month = {aug},
number = {8},
pages = {1845},
publisher = {Society of Photo-Optical Instrumentation Engineers},
title = {{Hyperparameter estimation in image restoration problems with partially known blurs}},
url = {http://opticalengineering.spiedigitallibrary.org/article.aspx?doi=10.1117/1.1487850},
volume = {41},
year = {2002}
}

This work is motivated by the observation that it is not possible to reliably estimate simultaneously all the necessary hyperparameters in an image restoration problem when the point-spread function is assumed to be the sum of a known deterministic and an unknown random component. To solve this problem we propose to use gamma hyperpriors for the unknown hyperparameters. Two iterative algorithms that simultaneously restore the image and estimate the hyperparameters are derived, based on the application of evidence analysis within the hierarchical Bayesian framework. Numerical experiments are presented that show the benefits of introducing hyperpriors for this problem. © 2002 Society of Photo-Optical Instrumentation Engineers.

@article{Lisimachos2001a,
abstract = {In this paper, we introduce a new methodology for single pass signal-to-noise ratio (SNR) video scalability based on the partitioning of the DCT coefficients. The DCT coefficients of the displaced frame difference (DFD) for inter-blocks or the intensity for intra-blocks are partitioned into a base layer and one or more enhancement layers, thus, producing an embedded bitstream. Subsets of this bitstream can be transmitted with increasing video quality as measured by the SNR. Given a bit budget for the base and enhancement layers the partitioning of the DCT coefficients is done in a way that is optimal in the operational rate-distortion sense. The optimization is performed using Lagrangian relaxation and dynamic programming (DP). Experimental results are presented and conclusions are drawn.},
author = {Kondi, L.P. and Katsaggelos, A.K.},
doi = {10.1109/83.967389},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
keywords = {Layered coding,Operational rate-distortion theory,Scalable video coding},
number = {11},
pages = {1613--1620},
title = {{An operational rate-distortion optimal single-pass SNR scalable video coder}},
url = {http://ieeexplore.ieee.org/document/967389/},
volume = {10},
year = {2001}
}

In this paper, we introduce a new methodology for single pass signal-to-noise ratio (SNR) video scalability based on the partitioning of the DCT coefficients. The DCT coefficients of the displaced frame difference (DFD) for inter-blocks or the intensity for intra-blocks are partitioned into a base layer and one or more enhancement layers, thus, producing an embedded bitstream. Subsets of this bitstream can be transmitted with increasing video quality as measured by the SNR. Given a bit budget for the base and enhancement layers the partitioning of the DCT coefficients is done in a way that is optimal in the operational rate-distortion sense. The optimization is performed using Lagrangian relaxation and dynamic programming (DP). Experimental results are presented and conclusions are drawn.

@article{tom2001high,
author = {Tom, B C and Galatsanos, N P and Katsaggelos, A K},
journal = {KLUWER INTERNATIONAL SERIES IN ENGINEERING AND COMPUTER SCIENCE},
pages = {73--106},
publisher = {KLUWER ACADEMIC PUBLISHERS GROUP},
title = {{High Resolution Image from Low Resolution Images}},
year = {2001}
}

@article{Aggelos2001a,
author = {Katsaggelos, A.K.},
doi = {10.1109/MSP.2001.924881},
issn = {1053-5888},
journal = {IEEE Signal Processing Magazine},
month = {may},
number = {3},
pages = {2--5},
title = {{Getting into vitual reality}},
url = {https://ieeexplore.ieee.org/document/924881/},
volume = {18},
year = {2001}
}

@article{lopez2001spect,
author = {Lopez, A. and Molina, R and Katsaggelos, A.K. and Mateos, J},
doi = {10.1109/ICASSP.2001.941318},
isbn = {0-7803-7041-4},
journal = {2001 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.01CH37221)},
number = {6-7},
pages = {1909--1912},
publisher = {IEEE},
title = {{SPECT image reconstruction using compound models}},
url = {http://ieeexplore.ieee.org/document/941318/},
volume = {3},
year = {2001}
}

@article{nygaard2001rate,
abstract = {Signal compression is an important problem encountered in many applications. Various techniques have been proposed over the years for addressing the problem. In this paper, we present a time domain algorithm based on the coding of line segments which are used to approximate the signal. These segments are fit in a way that is optimal in the rate distortion sense. Although the approach is applicable to any type of signal, we focus, in this paper, on the compression of electrocardiogram (ECG) signals. ECG signal compression has traditionally been tackled by heuristic approaches. However, it has been demonstrated [1] that exact optimization algorithms outperform these heuristic approaches by a wide margin with respect to reconstruction error. By formulating the compression problem as a graph theory problem, known optimization theory can be applied in order to yield optimal compression. In this paper, we present an algorithm that will guarantee the smallest possible distortion among all methods applying linear interpolation given an upper bound on the available number of bits. Using a varied signal test set, extensive coding experiments are presented. We compare the results from our coding method to traditional time domain ECG compression methods, as well as, to more recently developed frequency domain methods. Evaluation is based both on percentage root-mean-square difference (PRD) performance measure and visual inspection of the reconstructed signals. The results demonstrate that the exact optimization methods have superior performance compared to both traditional ECG compression methods and the frequency domain methods.},
author = {Nygaard, R. and Melnikov, G. and Katsaggelos, A.K.},
doi = {10.1109/10.900246},
issn = {00189294},
journal = {IEEE Transactions on Biomedical Engineering},
keywords = {Compression,ECG,Rate-distortion optimization,Shortest path},
number = {1},
pages = {28--40},
pmid = {11235588},
publisher = {IEEE},
title = {{A rate distortion optimal ECG coding algorithm}},
url = {http://ieeexplore.ieee.org/document/900246/},
volume = {48},
year = {2001}
}

Signal compression is an important problem encountered in many applications. Various techniques have been proposed over the years for addressing the problem. In this paper, we present a time domain algorithm based on the coding of line segments which are used to approximate the signal. These segments are fit in a way that is optimal in the rate distortion sense. Although the approach is applicable to any type of signal, we focus, in this paper, on the compression of electrocardiogram (ECG) signals. ECG signal compression has traditionally been tackled by heuristic approaches. However, it has been demonstrated [1] that exact optimization algorithms outperform these heuristic approaches by a wide margin with respect to reconstruction error. By formulating the compression problem as a graph theory problem, known optimization theory can be applied in order to yield optimal compression. In this paper, we present an algorithm that will guarantee the smallest possible distortion among all methods applying linear interpolation given an upper bound on the available number of bits. Using a varied signal test set, extensive coding experiments are presented. We compare the results from our coding method to traditional time domain ECG compression methods, as well as, to more recently developed frequency domain methods. Evaluation is based both on percentage root-mean-square difference (PRD) performance measure and visual inspection of the reconstructed signals. The results demonstrate that the exact optimization methods have superior performance compared to both traditional ECG compression methods and the frequency domain methods.

@article{tom2001resolution,
abstract = {In this paper, we propose an iterative algorithm for enhancing the resolution of monochrome and color image sequences. Various approaches toward motion estimation are investigated and compared. Improving the spatial resolution of an image sequence critically depends upon the accuracy of the motion estimator. The problem is complicated by the fact that the motion field is prone to significant errors since the original high-resolution images are not available. Improved motion estimates may be obtained by using a more robust and accurate motion estimator, such as a pel-recursive scheme instead of block matching. In processing color image sequences, there is the added advantage of having more flexibility in how the final motion estimates are obtained, and further improvement in the accuracy of the motion field is therefore possible. This is because there are three different intensity fields (channels) conveying the same motion information. In this paper, the choice of which motion estimator to use versus how the final estimates are obtained is weighed to see which issue is more critical in improving the estimated high-resolution sequences. Toward this end, an iterative algorithm is proposed, and two sets of experiments are presented. First, several different experiments using the same motion estimator but three different data fusion approaches to merge the individual motion fields were performed. Second, estimated high-resolution images using the block matching estimator were compared to those obtained by employing a pel recursive scheme. Experiments were performed on a real color image sequence, and performance was measured by the peak signal to noise ratio (PSNR).},
author = {Tom, B.C. and Katsaggelos, A.K.},
doi = {10.1109/83.902292},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
number = {2},
pages = {278--287},
publisher = {IEEE},
title = {{Resolution enhancement of monochrome and color video using motion compensation}},
url = {http://ieeexplore.ieee.org/document/902292/},
volume = {10},
year = {2001}
}

In this paper, we propose an iterative algorithm for enhancing the resolution of monochrome and color image sequences. Various approaches toward motion estimation are investigated and compared. Improving the spatial resolution of an image sequence critically depends upon the accuracy of the motion estimator. The problem is complicated by the fact that the motion field is prone to significant errors since the original high-resolution images are not available. Improved motion estimates may be obtained by using a more robust and accurate motion estimator, such as a pel-recursive scheme instead of block matching. In processing color image sequences, there is the added advantage of having more flexibility in how the final motion estimates are obtained, and further improvement in the accuracy of the motion field is therefore possible. This is because there are three different intensity fields (channels) conveying the same motion information. In this paper, the choice of which motion estimator to use versus how the final estimates are obtained is weighed to see which issue is more critical in improving the estimated high-resolution sequences. Toward this end, an iterative algorithm is proposed, and two sets of experiments are presented. First, several different experiments using the same motion estimator but three different data fusion approaches to merge the individual motion fields were performed. Second, estimated high-resolution images using the block matching estimator were compared to those obtained by employing a pel recursive scheme. Experiments were performed on a real color image sequence, and performance was measured by the peak signal to noise ratio (PSNR).

@article{mateos2000nonlinear,
abstract = {With block-based compression approaches for both still images and sequences of images annoying blocking artifacts are exhibited, primarily at high compression ratios. They are due to the independent processing (quantization) of the block transformed values of the intensity or the displaced frame difference. In this paper, we propose the application of the hierarchical Bayesian paradigm to the reconstruction of block discrete cosine transform (BDCT) compressed images and the estimation of the required parameters. We derive expressions for the iterative evaluation of these parameters applying the evidence analysis within the hierarchical Bayesian paradigm. The proposed method allows for the combination of parameters estimated at the coder and decoder. The performance of the proposed algorithms is demonstrated experimentally. {\textcopyright} 2000 IEEE.},
author = {Mateos, Javier and Katsaggelos, A.K. and Molina, Rafael},
doi = {10.1109/83.847833},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
keywords = {Bayesian models,Evidence analysis,Image coding,Post-processing,Reconstruction},
month = {jul},
number = {7},
pages = {1200--1215},
publisher = {New York, NY: Institute of Electrical and Electronics Engineers, 1992-},
title = {{A Bayesian approach for the estimation and transmission of regularization parameters for reducing blocking artifacts}},
url = {http://ieeexplore.ieee.org/document/847833/},
volume = {9},
year = {2000}
}

With block-based compression approaches for both still images and sequences of images annoying blocking artifacts are exhibited, primarily at high compression ratios. They are due to the independent processing (quantization) of the block transformed values of the intensity or the displaced frame difference. In this paper, we propose the application of the hierarchical Bayesian paradigm to the reconstruction of block discrete cosine transform (BDCT) compressed images and the estimation of the required parameters. We derive expressions for the iterative evaluation of these parameters applying the evidence analysis within the hierarchical Bayesian paradigm. The proposed method allows for the combination of parameters estimated at the coder and decoder. The performance of the proposed algorithms is demonstrated experimentally. © 2000 IEEE.

@article{Chun-Jen2000,
abstract = {Binocular camera systems are commonly used to construct 3-D-based scene description. However, there is a tradeoff between the length of the camera baseline and the difficulty of the matching problem and the extent of the field of view of the 3-D scene. A large baseline system provides better depth resolution than a smaller baseline system at the expense of a narrower field of view. To increase the depth resolution without increasing the difficulty of the matching problem and decreasing the field of view of the 3-D scene, a sequential 3-D based scene description technique is proposed in this paper. Multiple small-baseline 3-D scene descriptions from a single moving camera or an array of cameras are used to sequentially construct a large baseline 3-D scene description while maintaining the field of view of a small-baseline system. A Bayesian framework using a disparity-space image (DSI) technique for disparity estimation is presented. The cost function for large baseline image matching is designed based not only on the photometric matching error, the smoothness constraint, and the ordering constraint, but also on the previous disparity estimates from smaller baseline stereo image pairs as a prior model. Texture information is registered along the scan path of the camera(s). Experimental results demonstrate the effectiveness of this technique in visual communication applications.},
author = {{Chun-Jen Tsai} and Katsaggelos, Aggelos K A.K. and Tsai, Chun-Jen and Katsaggelos, Aggelos K A.K.},
doi = {10.1109/76.845002},
issn = {10518215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
month = {jun},
number = {4},
pages = {576--584},
publisher = {IEEE},
title = {{Sequential construction of 3-D-based scene description}},
url = {http://ieeexplore.ieee.org/document/845002/},
volume = {10},
year = {2000}
}

Binocular camera systems are commonly used to construct 3-D-based scene description. However, there is a tradeoff between the length of the camera baseline and the difficulty of the matching problem and the extent of the field of view of the 3-D scene. A large baseline system provides better depth resolution than a smaller baseline system at the expense of a narrower field of view. To increase the depth resolution without increasing the difficulty of the matching problem and decreasing the field of view of the 3-D scene, a sequential 3-D based scene description technique is proposed in this paper. Multiple small-baseline 3-D scene descriptions from a single moving camera or an array of cameras are used to sequentially construct a large baseline 3-D scene description while maintaining the field of view of a small-baseline system. A Bayesian framework using a disparity-space image (DSI) technique for disparity estimation is presented. The cost function for large baseline image matching is designed based not only on the photometric matching error, the smoothness constraint, and the ordering constraint, but also on the previous disparity estimates from smaller baseline stereo image pairs as a prior model. Texture information is registered along the scan path of the camera(s). Experimental results demonstrate the effectiveness of this technique in visual communication applications.

@article{Vladimir1998,
abstract = {In this paper, we examine the restoration problem when the point-spread function (PSF) of the degradation system is partially known. For this problem, the PSF is assumed to be the sum of a known deterministic and an unknown random component. This problem has been examined before; however, in most previous works the problem of estimating the parameters that define the restoration filters was not addressed. In this paper, two iterative algorithms that simultaneously restore the image and estimate the parameters of the restoration filter are proposed using evidence analysis (EA) within the hierarchical Bayesian framework. We show that the restoration step of the first of these algorithms is in effect almost identical to the regularized constrained total leastsquares (RCTLS) filter, while the restoration step of the second is identical to the linear minimum mean square-error (LMMSE) filter for this problem. Therefore, in this paper we provide a solution to the parameter estimation problem of the RCTLS filter. We further provide an alternative approach to the expectation-maximization (EM) framework to derive a parameter estimation algorithm for the LMMSE filter. These iterative algorithms are derived in the discrete Fourier transform (DFT) domain; therefore, they are computationally efficient even for large images. Numerical experiments are presented that test and compare the proposed algorithms.},
author = {Galatsanos, N.P. and Mesarovic, V.Z. and Molina, Rafael and Katsaggelos, A.K.},
doi = {10.1109/83.869189},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
number = {10},
pages = {1784--1797},
title = {{Hierarchical Bayesian image restoration from partially known blurs}},
url = {http://ieeexplore.ieee.org/document/869189/},
volume = {9},
year = {2000}
}

In this paper, we examine the restoration problem when the point-spread function (PSF) of the degradation system is partially known. For this problem, the PSF is assumed to be the sum of a known deterministic and an unknown random component. This problem has been examined before; however, in most previous works the problem of estimating the parameters that define the restoration filters was not addressed. In this paper, two iterative algorithms that simultaneously restore the image and estimate the parameters of the restoration filter are proposed using evidence analysis (EA) within the hierarchical Bayesian framework. We show that the restoration step of the first of these algorithms is in effect almost identical to the regularized constrained total leastsquares (RCTLS) filter, while the restoration step of the second is identical to the linear minimum mean square-error (LMMSE) filter for this problem. Therefore, in this paper we provide a solution to the parameter estimation problem of the RCTLS filter. We further provide an alternative approach to the expectation-maximization (EM) framework to derive a parameter estimation algorithm for the LMMSE filter. These iterative algorithms are derived in the discrete Fourier transform (DFT) domain; therefore, they are computationally efficient even for large images. Numerical experiments are presented that test and compare the proposed algorithms.

@article{Fabian2000,
abstract = {A major problem in object-oriented video coding is the efficient encoding of the shape information of arbitrarily shaped objects. Efficient shape coding schemes are also needed in encoding the shape information of video object (VO) in the upcoming MPEG-4 standard. Furthermore, there are many applications where only the shape needs to be encoded, such as CAD, 3D modeling and signature encoding. In this paper, we present an efficient method for the lossy encoding of object shapes which are given as 8-connect chain codes using a mathematical model. We approximate a boundary by a second-order B-spline curve and consider the problem of finding the curve with the lowest bit-rate for a given distortion. The presented scheme is optimal, efficient and offers complete control over the trade-off between bit-rate and distortion. It is an extension of our previous research where we used polygons to approximate a boundary. The main reason for using curves rather than polygons is that curves have a more natural appearance than polygons and can give better coding efficiencies. We present results of the proposed scheme using objects boundaries in different shapes and sizes as well as an MPEG-4 test sequence.},
author = {Meier, Fabian W. and Schuster, Guido M. and Katsaggelos, Aggelos K.},
doi = {10.1016/S0923-5965(99)00045-4},
issn = {09235965},
journal = {Signal Processing: Image Communication},
month = {may},
number = {7-8},
pages = {685--701},
title = {{A mathematical model for shape coding with B-splines}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0923596599000454},
volume = {15},
year = {2000}
}

A major problem in object-oriented video coding is the efficient encoding of the shape information of arbitrarily shaped objects. Efficient shape coding schemes are also needed in encoding the shape information of video object (VO) in the upcoming MPEG-4 standard. Furthermore, there are many applications where only the shape needs to be encoded, such as CAD, 3D modeling and signature encoding. In this paper, we present an efficient method for the lossy encoding of object shapes which are given as 8-connect chain codes using a mathematical model. We approximate a boundary by a second-order B-spline curve and consider the problem of finding the curve with the lowest bit-rate for a given distortion. The presented scheme is optimal, efficient and offers complete control over the trade-off between bit-rate and distortion. It is an extension of our previous research where we used polygons to approximate a boundary. The main reason for using curves rather than polygons is that curves have a more natural appearance than polygons and can give better coding efficiencies. We present results of the proposed scheme using objects boundaries in different shapes and sizes as well as an MPEG-4 test sequence.

@article{Yao2000,
abstract = {Encoder error resilient (ER) mechanisms adopted by coding standards, and decoder-based error concealment techniques and some closed loop methods are discussed. These techniques provide a complete overview of low latency, low bit-rate video transmission methods for error prone environments. It is found that video coding standards, H.263 and MPEG-4 by including a wide variety of ER tools, can lead to acceptable video quality in highly error prone environment.},
author = {Wang, Yao and Wenger, Stephan and Wen, Jiangtao and Katsaggelos, Aggelos K. A.K. and {Yao Wang} and Wenger, Stephan and {Jiantao Wen} and Katsaggelos, Aggelos K. A.K.},
chapter = {61},
doi = {10.1109/79.855913},
isbn = {10535888},
issn = {10535888},
journal = {IEEE Signal Processing Magazine},
month = {jul},
number = {4},
pages = {61--82},
publisher = {IEEE},
title = {{Error resilient video coding techniques}},
url = {http://ieeexplore.ieee.org/document/855913/},
volume = {17},
year = {2000}
}

Encoder error resilient (ER) mechanisms adopted by coding standards, and decoder-based error concealment techniques and some closed loop methods are discussed. These techniques provide a complete overview of low latency, low bit-rate video transmission methods for error prone environments. It is found that video coding standards, H.263 and MPEG-4 by including a wide variety of ER tools, can lead to acceptable video quality in highly error prone environment.

@article{Gerry2000,
abstract = {This paper investigates ways to explore the between frame correlation of shape information within the framework of an operationally rate-distortion (ORD) optimized coder. Contours are approximated both by connected second-order spline segments, each defined by three consecutive control points, and by segments of the motion-compensated reference contours. Consecutive control points are then encoded predictively using angle and run temporal contexts or by tracking the reference contour. We utilize a novel criterion for selecting global object motion vectors, which improves efficiency. The problem is formulated as Lagrangian minimization and solved using dynamic programming. Furthermore, we employ an iterative technique to remove dependency on a particular variable length code and jointly arrive at the ORD globally optimal solution and an optimized conditional parameter distribution.},
author = {Melnikov, Gerry and Schuster, G.M. and Katsaggelos, A.K.},
doi = {10.1109/76.856451},
issn = {10518215},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
number = {5},
pages = {744--754},
title = {{Shape coding using temporal correlation and joint VLC optimization}},
url = {http://ieeexplore.ieee.org/document/856451/},
volume = {10},
year = {2000}
}

This paper investigates ways to explore the between frame correlation of shape information within the framework of an operationally rate-distortion (ORD) optimized coder. Contours are approximated both by connected second-order spline segments, each defined by three consecutive control points, and by segments of the motion-compensated reference contours. Consecutive control points are then encoded predictively using angle and run temporal contexts or by tracking the reference contour. We utilize a novel criterion for selecting global object motion vectors, which improves efficiency. The problem is formulated as Lagrangian minimization and solved using dynamic programming. Furthermore, we employ an iterative technique to remove dependency on a particular variable length code and jointly arrive at the ORD globally optimal solution and an optimized conditional parameter distribution.

@article{Laura1999,
author = {Drake, Laura A. and Rutledge, Janet C. and Katsaggelos, Aggelos},
doi = {10.1121/1.427622},
issn = {0001-4966},
journal = {The Journal of the Acoustical Society of America},
month = {oct},
number = {4},
pages = {2238--2238},
title = {{Computational auditory scene analysis‐constrained array processing for sound source separation}},
url = {http://asa.scitation.org/doi/10.1121/1.427622},
volume = {106},
year = {1999}
}

@article{Schuster1999,
author = {Schuster, G.M. and Melnikov, G. and Katsaggelos, A.K.},
doi = {10.1109/6046.748167},
issn = {15209210},
journal = {IEEE Transactions on Multimedia},
month = {mar},
number = {1},
pages = {3--17},
title = {{A review of the minimum maximum criterion for optimal bit allocation among dependent quantizers}},
url = {http://ieeexplore.ieee.org/document/748167/},
volume = {1},
year = {1999}
}

@article{Laura1999a,
author = {Drake, Laura A. and Katsaggelos, Aggelos and Zhang, Jun},
doi = {10.1121/1.427787},
issn = {0001-4966},
journal = {The Journal of the Acoustical Society of America},
month = {oct},
number = {4},
pages = {2277--2278},
title = {{Normalizing inter‐band scaling in minimum variance (MV) beamformer estimates of variable‐bandwidth filtered wideband signals}},
url = {http://asa.scitation.org/doi/10.1121/1.427787},
volume = {106},
year = {1999}
}

@article{bandyopadhyay2001video,
abstract = {Although the transmission of video images over telephone lines has a long history, new technology, particularly video compression techniques, now makes video telephony affordable. The technology is practical, and it can be successful if systems maintain high quality and ease of use. The results of behavioral science studies of visual communication indicate that adding video to voice calls increases the effectiveness of human communication. Human factors and behavioral science techniques can evaluate the quality of visual communication systems and make such systems easy for customers to use. {\textcopyright} 1993 AT&T Technical Journal},
address = {Hoboken, NJ, USA},
author = {Bandyopadhyay, Saurav K. and Kondi, Lisimachos P. and Schuster, Guido M. and Katsaggelos, Aggelos K.},
doi = {10.1002/047134608X.W1919},
issn = {15387305},
journal = {Wiley Encyclopedia of Electrical and Electronics Engineering},
month = {dec},
number = {3},
pages = {7--20},
publisher = {John Wiley & Sons, Inc.},
title = {{Video Telephony}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/047134608X.W1919},
volume = {72},
year = {1999}
}

Although the transmission of video images over telephone lines has a long history, new technology, particularly video compression techniques, now makes video telephony affordable. The technology is practical, and it can be successful if systems maintain high quality and ease of use. The results of behavioral science studies of visual communication indicate that adding video to voice calls increases the effectiveness of human communication. Human factors and behavioral science techniques can evaluate the quality of visual communication systems and make such systems easy for customers to use. © 1993 AT&T Technical Journal

@article{Aggelosf,
abstract = {Many optical flow estimation techniques are based on the differential optical flow equation. For these techniques, a locally constant flow model is typically used to allow the construction of an over-determined system of constraint equations. In this paper, the problem of solving the system of optical flow equations using a constraint total least squares (CTLS) approach is investigated. It is shown that by modifying the CTLS approach it becomes identical to a maximum likelihood (ML) approach to the problem. This modification improves the CTLS estimates especially when the estimation window size is small, as is demonstrated experimentally.},
author = {Tsai, C and Galatsanos, N P and Katsaggelos, A K},
journal = {Nsip},
keywords = {&account &address &adjacency &age &composed &const},
pages = {53--56},
title = {{Maximum-likelihood optical flow estimation using differential constraints}},
year = {1999}
}

Many optical flow estimation techniques are based on the differential optical flow equation. For these techniques, a locally constant flow model is typically used to allow the construction of an over-determined system of constraint equations. In this paper, the problem of solving the system of optical flow equations using a constraint total least squares (CTLS) approach is investigated. It is shown that by modifying the CTLS approach it becomes identical to a maximum likelihood (ML) approach to the problem. This modification improves the CTLS estimates especially when the estimation window size is small, as is demonstrated experimentally.

@article{Hong1999,
author = {Hong, Min-Cheol and Schwab, Harald and Kondi, Lisimachos P. and Katsaggelos, Aggelos K.},
doi = {10.1016/S0923-5965(98)00061-7},
issn = {09235965},
journal = {Signal Processing: Image Communication},
month = {may},
number = {6-8},
pages = {473--492},
title = {{Error concealment algorithms for compressed video}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0923596598000617},
volume = {14},
year = {1999}
}

@article{Aggelos1999,
author = {Katsaggelos, A.K. and Jamieson, L.H.},
doi = {10.1109/MSP.1999.799931},
issn = {1053-5888},
journal = {IEEE Signal Processing Magazine},
month = {nov},
number = {6},
pages = {2--6},
title = {{So Simple evan a child can do it}},
url = {http://ieeexplore.ieee.org/document/799931/},
volume = {16},
year = {1999}
}

@article{molina1999bayesian,
abstract = {In this paper, we propose the application of the hierarchical Bayesian paradigm to the image restoration problem. We derive expressions for the iterative evaluation of the two hyperparameters applying the evidence and maximum a posteriori (MAP) analysis within the hierarchical Bayesian paradigm. We show analytically that the analysis provided by the evidence approach is more realistic and appropriate than the MAP approach for the image restoration problem. We furthermore study the relationship between the evidence and an iterative approach resulting from the set theoretic regularization approach for estimating the two hyperparameters, or their ratio, defined as the regularization parameter. Finally the proposed algorithms are tested experimentally. {\textcopyright} 1999 IEEE.},
author = {Molina, Rafael and Katsaggelos, A.K. and Mateos, Javier},
doi = {10.1109/83.743857},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
keywords = {Hierarchical bayesian models,Image restoration,Parameter estimation,Regularization},
number = {2},
pages = {231--246},
publisher = {IEEE},
title = {{Bayesian and regularization methods for hyperparameter estimation in image restoration}},
url = {http://ieeexplore.ieee.org/document/743857/},
volume = {8},
year = {1999}
}

In this paper, we propose the application of the hierarchical Bayesian paradigm to the image restoration problem. We derive expressions for the iterative evaluation of the two hyperparameters applying the evidence and maximum a posteriori (MAP) analysis within the hierarchical Bayesian paradigm. We show analytically that the analysis provided by the evidence approach is more realistic and appropriate than the MAP approach for the image restoration problem. We furthermore study the relationship between the evidence and an iterative approach resulting from the set theoretic regularization approach for estimating the two hyperparameters, or their ratio, defined as the regularization parameter. Finally the proposed algorithms are tested experimentally. © 1999 IEEE.

@article{Chun-Jen1999c,
abstract = {A new divide-and-conquer technique for disparity estimation is proposed in this paper. This technique performs feature matching following the high confidence first principle, starting with the strongest feature point in the stereo pair of scanlines. Once the first matching pair is established, the ordering constraint in disparity estimation allows the original intra-scanline matching problem to be divided into two smaller subproblems. Each subproblem can then be solved recursively until there is no reliable feature point within the subintervals. This technique is very efficient for dense disparity map estimation for stereo images with rich features. For general scenes, this technique can be paired up with the disparity-space image (DSI) technique to compute dense disparity maps with integrated occlusion detection. In this approach, the divide-and-conquer part of the algorithm handles the matching of stronger features and the DSI-based technique handles the matching of pixels in between feature points and the detection of occlusions. An extension to the standard disparity-space technique is also presented to compliment the divide-and-conquer algorithm. Experiments demonstrate the effectiveness of the proposed divide-and-conquer DSI algorithm. {\textcopyright} 1999 IEEE.},
author = {Tsai, Chun Jen and Katsaggelos, Aggelos K. A.K. and {Chun-Jen Tsai} and Katsaggelos, Aggelos K. A.K.},
chapter = {18},
doi = {10.1109/6046.748168},
isbn = {15209210},
issn = {15209210},
journal = {IEEE Transactions on Multimedia},
keywords = {Depth map,Disparity estimation,Disparity-space image,Divide-and-conquer stereo matching},
month = {mar},
number = {1},
pages = {18--29},
publisher = {IEEE},
title = {{Dense disparity estimation with a divide-and-conquer disparity space image technique}},
url = {http://ieeexplore.ieee.org/document/748168/},
volume = {1},
year = {1999}
}

A new divide-and-conquer technique for disparity estimation is proposed in this paper. This technique performs feature matching following the high confidence first principle, starting with the strongest feature point in the stereo pair of scanlines. Once the first matching pair is established, the ordering constraint in disparity estimation allows the original intra-scanline matching problem to be divided into two smaller subproblems. Each subproblem can then be solved recursively until there is no reliable feature point within the subintervals. This technique is very efficient for dense disparity map estimation for stereo images with rich features. For general scenes, this technique can be paired up with the disparity-space image (DSI) technique to compute dense disparity maps with integrated occlusion detection. In this approach, the divide-and-conquer part of the algorithm handles the matching of stronger features and the DSI-based technique handles the matching of pixels in between feature points and the detection of occlusions. An extension to the standard disparity-space technique is also presented to compliment the divide-and-conquer algorithm. Experiments demonstrate the effectiveness of the proposed divide-and-conquer DSI algorithm. © 1999 IEEE.

@article{Brian1998,
abstract = {In this article, the problem of detecting and removing anomalies in digitized animation film is addressed. The impetus of this article comes from the motion picture industry, where several studios are rereleasing vintage film to the public which are often accompanied by visual degradation in the film color, grain, and overall quality. These degradations, or anomalies, can be seen as unwanted visual artifacts that are usually larger than one pixel in size and appear in only one frame. One of the novelties of this article is that the restoration of animation film, as opposed to "real-world" image sequences, is investigated. It often contains additional artifacts and raises other issues not found in most sequences. These artifacts are first defined, and the appropriate steps for their detection and removal are described. Experiments with the proposed algorithm were performed using scenes from the animation film Fantasia, courtesy of Walt Disney Feature Animations, and are discussed in detail. Although our algorithm was developed for the removal of artifacts on animation film, aspects of it can be applied to nonanimated film. {\textcopyright} 1998 John Wiley & Sons, Inc.},
author = {Tom, Brian C. and Kang, Moon Gi and Hong, Min-Cheol and Katsaggelos, Aggelos K.},
doi = {10.1002/(SICI)1098-1098(1998)9:4<283::AID-IMA10>3.0.CO;2-Y},
issn = {0899-9457},
journal = {International Journal of Imaging Systems and Technology},
number = {4},
pages = {283--293},
title = {{Detection and removal of anomalies in digitized animation film}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/(SICI)1098-1098(1998)9:4%3C283::AID-IMA10%3E3.0.CO;2-Y},
volume = {9},
year = {1998}
}

In this article, the problem of detecting and removing anomalies in digitized animation film is addressed. The impetus of this article comes from the motion picture industry, where several studios are rereleasing vintage film to the public which are often accompanied by visual degradation in the film color, grain, and overall quality. These degradations, or anomalies, can be seen as unwanted visual artifacts that are usually larger than one pixel in size and appear in only one frame. One of the novelties of this article is that the restoration of animation film, as opposed to "real-world" image sequences, is investigated. It often contains additional artifacts and raises other issues not found in most sequences. These artifacts are first defined, and the appropriate steps for their detection and removal are described. Experiments with the proposed algorithm were performed using scenes from the animation film Fantasia, courtesy of Walt Disney Feature Animations, and are discussed in detail. Although our algorithm was developed for the removal of artifacts on animation film, aspects of it can be applied to nonanimated film. © 1998 John Wiley & Sons, Inc.

@article{Chellappa1998,
author = {Chellappa, R and Fukushima, K and Katsaggelos, A.K. and {Sun-Yuan Kung} and LeCun, Y and Nasrabadi, N.M. and Poggio, T.A.},
chapter = {1093},
doi = {10.1109/TIP.1998.704303},
isbn = {1057-7149 1941-0042},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {aug},
number = {8},
pages = {1093--1096},
title = {{Guest Editorial Applications Of Artificial Neural Networks To Image Processing}},
url = {http://ieeexplore.ieee.org/document/704303/},
volume = {7},
year = {1998}
}

@article{schuster1998optimal,
abstract = {In this paper, we propose an optimal quadtree (QT)-based motion estimator for video compression. It is optimal in the sense that for a given bit budget for encoding the displacement vector field (DVF) and the QT segmentation, the scheme finds a DVF and a QT segmentation which minimizes the energy of the resulting displaced frame difference (DFD). We find the optimal QT decomposition and the optimal DVF jointly using the Lagrangian multiplier method and a multilevel dynamic program. We introduce a new, very fast convex search for the optimal Lagrangian multiplier $\lambda$*, which results in a very fast convergence of the Lagrangian multiplier method. The resulting DVF is spatially inhomogeneous, since large blocks are used in areas with simple motion and small blocks in areas with complex motion. We also propose a novel motion-compensated interpolation scheme which uses the same mathematical tools developed for the QT-based motion estimator. One of the advantages of this scheme is the globally optimal control of the tradeoff between the interpolation error energy and the DVF smoothness. Another advantage is that no interpolation of the DVF is required since we directly estimate the DVF and the QT-segmentation for the frame which needs to be interpolated. We present results with the proposed QT-based motion estimator which show that for the same DFD energy the proposed estimator uses about 25% fewer bits than the commonly used block matching algorithm. We also experimentally compare the interpolated frames using the proposed motion compensated interpolation scheme with the reconstructed original frames. This comparison demonstrates the effectiveness of the proposed interpolation scheme. {\textcopyright} 1998 IEEE.},
author = {Schuster, G.M. and Katsaggelos, A.K.},
doi = {10.1109/83.725359},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
keywords = {Dynamic programming,Lagrangian relaxation,Motion estimation,Motion-compensated interpolation,Optimal bit allocation,Rate distortion theory,Video compression},
number = {11},
pages = {1505--1523},
publisher = {IEEE},
title = {{An optimal quadtree-based motion estimation and motion-compensated interpolation scheme for video compression}},
url = {http://ieeexplore.ieee.org/document/725359/},
volume = {7},
year = {1998}
}

In this paper, we propose an optimal quadtree (QT)-based motion estimator for video compression. It is optimal in the sense that for a given bit budget for encoding the displacement vector field (DVF) and the QT segmentation, the scheme finds a DVF and a QT segmentation which minimizes the energy of the resulting displaced frame difference (DFD). We find the optimal QT decomposition and the optimal DVF jointly using the Lagrangian multiplier method and a multilevel dynamic program. We introduce a new, very fast convex search for the optimal Lagrangian multiplier $λ$*, which results in a very fast convergence of the Lagrangian multiplier method. The resulting DVF is spatially inhomogeneous, since large blocks are used in areas with simple motion and small blocks in areas with complex motion. We also propose a novel motion-compensated interpolation scheme which uses the same mathematical tools developed for the QT-based motion estimator. One of the advantages of this scheme is the globally optimal control of the tradeoff between the interpolation error energy and the DVF smoothness. Another advantage is that no interpolation of the DVF is required since we directly estimate the DVF and the QT-segmentation for the frame which needs to be interpolated. We present results with the proposed QT-based motion estimator which show that for the same DFD energy the proposed estimator uses about 25% fewer bits than the commonly used block matching algorithm. We also experimentally compare the interpolated frames using the proposed motion compensated interpolation scheme with the reconstructed original frames. This comparison demonstrates the effectiveness of the proposed interpolation scheme. © 1998 IEEE.

@article{haris1998hybrid,
abstract = {A hybrid multidimensional image segmentation algorithm is proposed, which combines edge and region-based techniques through the morphological algorithm of watersheds. An edge-preserving statistical noise reduction approach is used as a preprocessing stage in order to compute an accurate estimate of the image gradient. Then, an initial partitioning of the image into primitive regions is produced by applying the watershed transform on the image gradient magnitude. This initial segmentation is the input to a computationally efficient hierarchical (bottom-up) region merging process that produces the final segmentation. The latter process uses the region adjacency graph (RAG) representation of the image regions. At each step, the most similar pair of regions is determined (minimum cost RAG edge), the regions are merged and the RAG is updated. Traditionally, the above is implemented by storing all RAG edges in a priority queue. We propose a significantly faster algorithm, which additionally maintains the so-called nearest neighbor graph, due to which the priority queue size and processing time are drastically reduced. The final segmentation provides, due to the RAG, one-pixel wide, closed, and accurately localized contours/surfaces. Experimental results obtained with two-dimensional/three-dimensional (2-D/3-D) magnetic resonance images are presented. {\textcopyright} 1998 IEEE.},
author = {Haris, Kostas and Efstratiadis, S.N. and Maglaveras, Nicos and Katsaggelos, A.K.},
doi = {10.1109/83.730380},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
keywords = {Image segmentation,Nearest neighbor region merging,Noise reduction,Watershed transform},
number = {12},
pages = {1684--1699},
publisher = {IEEE},
title = {{Hybrid image segmentation using watersheds and fast region merging}},
url = {http://ieeexplore.ieee.org/document/730380/},
volume = {7},
year = {1998}
}

A hybrid multidimensional image segmentation algorithm is proposed, which combines edge and region-based techniques through the morphological algorithm of watersheds. An edge-preserving statistical noise reduction approach is used as a preprocessing stage in order to compute an accurate estimate of the image gradient. Then, an initial partitioning of the image into primitive regions is produced by applying the watershed transform on the image gradient magnitude. This initial segmentation is the input to a computationally efficient hierarchical (bottom-up) region merging process that produces the final segmentation. The latter process uses the region adjacency graph (RAG) representation of the image regions. At each step, the most similar pair of regions is determined (minimum cost RAG edge), the regions are merged and the RAG is updated. Traditionally, the above is implemented by storing all RAG edges in a priority queue. We propose a significantly faster algorithm, which additionally maintains the so-called nearest neighbor graph, due to which the priority queue size and processing time are drastically reduced. The final segmentation provides, due to the RAG, one-pixel wide, closed, and accurately localized contours/surfaces. Experimental results obtained with two-dimensional/three-dimensional (2-D/3-D) magnetic resonance images are presented. © 1998 IEEE.

@article{schuster1998optimal,
abstract = {In this paper, we present fast and efficient methods for the lossy encoding of object boundaries that are given as eight-connect chain codes. We approximate the boundary by a polygon, and consider the problem of finding the polygon which leads to the smallest distortion for a given number of bits. We also address the dual problem of finding the polygon which leads to the smallest bit rate for a given distortion. We consider two different classes of distortion measures. The first class is based on the maximum operator and the second class is based on the summation operator. For the first class, we derive a fast and optimal scheme that is based on a shortest path algorithm for a weighted directed acyclic graph. For the second class we propose a solution approach that is based on the Lagrange multiplier method, which uses the above-mentioned shortest path algorithm. Since the Lagrange multiplier method can only find solutions on the convex hull of the operational rate distortion function, we also propose a tree-pruning-based algorithm that can find all the optimal solutions. Finally, we present results of the proposed schemes using objects from the Miss America sequence. {\textcopyright} 1998 IEEE.},
author = {Schuster, G.M. and Katsaggelos, A.K.},
doi = {10.1109/83.650847},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
keywords = {Boundary encoding,Dynamic programming,Min-max optimization,Object-oriented video compression,Operational rate distortion theory,Shape encoding,Tree pruning},
number = {1},
pages = {13--26},
publisher = {IEEE},
title = {{An optimal polygonal boundary encoding scheme in the rate distortion sense}},
url = {http://ieeexplore.ieee.org/document/650847/},
volume = {7},
year = {1998}
}

In this paper, we present fast and efficient methods for the lossy encoding of object boundaries that are given as eight-connect chain codes. We approximate the boundary by a polygon, and consider the problem of finding the polygon which leads to the smallest distortion for a given number of bits. We also address the dual problem of finding the polygon which leads to the smallest bit rate for a given distortion. We consider two different classes of distortion measures. The first class is based on the maximum operator and the second class is based on the summation operator. For the first class, we derive a fast and optimal scheme that is based on a shortest path algorithm for a weighted directed acyclic graph. For the second class we propose a solution approach that is based on the Lagrange multiplier method, which uses the above-mentioned shortest path algorithm. Since the Lagrange multiplier method can only find solutions on the convex hull of the operational rate distortion function, we also propose a tree-pruning-based algorithm that can find all the optimal solutions. Finally, we present results of the proposed schemes using objects from the Miss America sequence. © 1998 IEEE.

@article{Guido1998,
abstract = {A framework for the rate-distortion operationally optimal encoding of shape information in the intra mode is presented. It is shown that each curve approximation has a natural order. If the control-point encoding scheme is matched to this order and the distortion is carefully defined, then the optimal approximation can be found using a directed acyclic graph (DAG)-shortest-path algorithm. The minimum-maximum distortion optimization problem and the minimum total (average) distortion optimization problem can be both solved by similar means, using an appropriate definition of the DAG weight function.},
author = {Schuster, G.M. and Melnikov, Gerry and Katsaggelos, A.K.},
doi = {10.1109/79.733498},
issn = {10535888},
journal = {IEEE Signal Processing Magazine},
number = {6},
pages = {91--108},
publisher = {IEEE},
title = {{Operationally optimal vertex-based shape coding}},
url = {http://ieeexplore.ieee.org/document/733498/},
volume = {15},
year = {1998}
}

A framework for the rate-distortion operationally optimal encoding of shape information in the intra mode is presented. It is shown that each curve approximation has a natural order. If the control-point encoding scheme is matched to this order and the distortion is carefully defined, then the optimal approximation can be found using a directed acyclic graph (DAG)-shortest-path algorithm. The minimum-maximum distortion optimization problem and the minimum total (average) distortion optimization problem can be both solved by similar means, using an appropriate definition of the DAG weight function.

@article{schuster1997video,
abstract = {In this paper, we present a theory for the optimal bit allocation among quadtree (QT) segmentation, displacement vector field (DVF), and displaced frame difference (DFD). The theory is applicable to variable block size motion-compensated video coders (VBSMCVC), where the variable block sizes are encoded using the QT structure, the DVF is encoded by first-order differential pulse code modulation (DPCM), the DFD is encoded by a block-based scheme, and an additive distortion measure is employed. We derive an optimal scanning path for a QT that is based on a Hilbert curve. We consider the case of a lossless VBSMCVC first, for which we develop the optimal bit allocation algorithm using dynamic programming (DP). We then consider a lossy VBSMCVC, for which we use Lagrangian relaxation, and show how an iterative scheme, which employs the DP-based solution, can be used to find the optimal solution. We finally present a VBSMCVC, which is based on the proposed theory, which employs a DCT-based DFD encoding scheme. We compare the proposed coder with H.263. The results show that it outperforms H.263 significantly in the rate distortion sense, as well as in the subjective sense. {\textcopyright} 1997 IEEE.},
author = {Schuster, G.M. and Katsaggelos, A.K.},
doi = {10.1109/83.641410},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
keywords = {Dynamic programming,Lagrangian relaxation,Optimal bit allocation,Rate distortion theory,Video compression},
month = {nov},
number = {11},
pages = {1487--1502},
publisher = {IEEE},
title = {{A video compression scheme with optimal bit allocation among segmentation, motion, and residual error}},
url = {https://ieeexplore.ieee.org/document/641410/},
volume = {6},
year = {1997}
}

In this paper, we present a theory for the optimal bit allocation among quadtree (QT) segmentation, displacement vector field (DVF), and displaced frame difference (DFD). The theory is applicable to variable block size motion-compensated video coders (VBSMCVC), where the variable block sizes are encoded using the QT structure, the DVF is encoded by first-order differential pulse code modulation (DPCM), the DFD is encoded by a block-based scheme, and an additive distortion measure is employed. We derive an optimal scanning path for a QT that is based on a Hilbert curve. We consider the case of a lossless VBSMCVC first, for which we develop the optimal bit allocation algorithm using dynamic programming (DP). We then consider a lossy VBSMCVC, for which we use Lagrangian relaxation, and show how an iterative scheme, which employs the DP-based solution, can be used to find the optimal solution. We finally present a VBSMCVC, which is based on the proposed theory, which employs a DCT-based DFD encoding scheme. We compare the proposed coder with H.263. The results show that it outperforms H.263 significantly in the rate distortion sense, as well as in the subjective sense. © 1997 IEEE.

@article{kang1997simultaneous,
abstract = {In this correspondence, a constrained least-squares multichannel image restoration approach is proposed, in which no prior knowledge of the noise variance at each channel or the degree of smoothness of the original image is required. The regularization functional for each channel is determined by incorporating both within-channel and cross-channel information. It is shown that the proposed smoothing functional has a global minimizer. {\textcopyright} 1997 IEEE.},
author = {{Moon Gi Kang} and Katsaggelos, Aggelos K. A.K. and Kang, Moon Gi and Katsaggelos, Aggelos K. A.K.},
doi = {10.1109/83.568936},
issn = {1057-7149},
journal = {IEEE Transactions on Image Processing},
month = {may},
number = {5},
pages = {774--778},
publisher = {IEEE},
title = {{Simultaneous multichannel image restoration and estimation of the regularization parameters}},
url = {https://ieeexplore.ieee.org/document/568936/},
volume = {6},
year = {1997}
}

In this correspondence, a constrained least-squares multichannel image restoration approach is proposed, in which no prior knowledge of the noise variance at each channel or the degree of smoothness of the original image is required. The regularization functional for each channel is determined by incorporating both within-channel and cross-channel information. It is shown that the proposed smoothing functional has a global minimizer. © 1997 IEEE.

@article{schuster1997theory,
abstract = {In this paper, we address the fundamental problem of optimally splitting a video sequence into two sources of information, the displaced frame difference (DFD) and the displacement vector field (DVF). We first consider the case of a lossless motion-compensated video coder (MCVC), and derive a general dynamic programming (DP) formulation which results in an optimal tradeoff between the DVF and the DFD. We then consider the more important case of a lossy MCVC, and present an algorithm which solves the tradeoff between the rate and the distortion. This algorithm is based on the Lagrange multiplier method and the DP approach introduced for the lossless MCVC. We then present an H.263-based MCVC which uses the proposed optimal bit allocation, and compare its results to H.263. As expected, the proposed coder is superior in the rate-distortion sense. In addition to this, it offers many advantages for a rate control scheme. The presented theory can be applied to build new optimal coders, and to analyze the heuristics employed in existing coders. In fact, whenever one changes an existing coder, the proposed theory can be used to evaluate how the change affects its performance.},
author = {Schuster, G.M. and Katsaggelos, A.K.},
doi = {10.1109/49.650047},
issn = {07338716},
journal = {IEEE Journal on Selected Areas in Communications},
keywords = {Motion estimation,Optimal bit allocation,Quantizer selection,Rate-distortion theory,Video compression},
number = {9},
pages = {1739--1751},
publisher = {IEEE},
title = {{A theory for the optimal bit allocation between displacement vector field and displaced frame difference}},
url = {http://ieeexplore.ieee.org/document/650047/},
volume = {15},
year = {1997}
}

In this paper, we address the fundamental problem of optimally splitting a video sequence into two sources of information, the displaced frame difference (DFD) and the displacement vector field (DVF). We first consider the case of a lossless motion-compensated video coder (MCVC), and derive a general dynamic programming (DP) formulation which results in an optimal tradeoff between the DVF and the DFD. We then consider the more important case of a lossy MCVC, and present an algorithm which solves the tradeoff between the rate and the distortion. This algorithm is based on the Lagrange multiplier method and the DP approach introduced for the lossless MCVC. We then present an H.263-based MCVC which uses the proposed optimal bit allocation, and compare its results to H.263. As expected, the proposed coder is superior in the rate-distortion sense. In addition to this, it offers many advantages for a rate control scheme. The presented theory can be applied to build new optimal coders, and to analyze the heuristics employed in existing coders. In fact, whenever one changes an existing coder, the proposed theory can be used to evaluate how the change affects its performance.

@article{banham1997digital,
author = {Banham, M.R. and Katsaggelos, A.K.},
doi = {10.1109/79.581363},
issn = {10535888},
journal = {IEEE Signal Processing Magazine},
month = {mar},
number = {2},
pages = {24--41},
publisher = {IEEE},
title = {{Digital image restoration}},
url = {http://ieeexplore.ieee.org/document/581363/},
volume = {14},
year = {1997}
}

@article{chellappa1997audio,
author = {Chellappa, Rama and {Tsuhan Chen} and Katsaggelos, A.},
doi = {10.1109/79.598590},
issn = {10535888},
journal = {IEEE Signal Processing Magazine},
month = {jul},
number = {4},
pages = {37--38},
publisher = {IEEE},
title = {{Audio-visual interaction in multimodal communication}},
url = {http://ieeexplore.ieee.org/document/598590/},
volume = {14},
year = {1997}
}

@article{woods1997rukf,
author = {Woods, J W and Hsiang, S T and Banham, M and Katsaggelos, A},
journal = {Ieee Signal Processing Magazine},
number = {6},
pages = {12},
title = {{RUKF performance revisited}},
volume = {14},
year = {1997}
}

@article{Taner1997c,
abstract = {There are a large number of applications requiring the compression of video at Very Low Bit Rates (VLBR). Such applications include wireless video conferencing, video over the internet, multimedia database retrieval and remote sensing and monitoring. Recently, the MPEG-4 standardization effort has been a motivating factor to find a solution to this challenging problem. The existing approaches to this problem can generally be grouped into block-based, model-based, and object-oriented. Block-based approaches follow the traditional strategy of decoupling the image sequence into blocks, model-based approaches rely on complex 3-D models for specific objects that are encoded, and object-oriented approaches rely on analyzing the scene into differently moving objects. All three approaches exhibit potential problems. Block-based approaches tend to generate artifacts at the boundaries of the blocks, as well as to limit the minimum achievable bit-rate due to the fixed analysis structure of the scene. Model-based codecs are limited by the complex 3-D models of the objects to be encoded. On the other hand, object-oriented codecs can generate a significant overhead due to the analysis of the scene which needs to be transmitted, which in turn can be the limiting factor in achieving the target bit-rates. In this paper, we propose a hybrid object-oriented codec in which the correlations among the three information fields, e.g., motion, segmentation and intensity fields, are exploited both spatially and temporally. In the proposed method, additional intelligence is given to the decoder, resulting in a reduction of the required bandwidth. The residual information is analyzed into three different categories, i.e., occlusion, model failures, and global refinement. The residual information is encoded and transmitted across the channel with other side information. Experimental results are presented which demonstrate the effectiveness of the proposed approach.},
author = {{\"{O}}zcelik, Taner and Katsaggelos, Aggelos K.},
doi = {https://doi.org/10.1023/A:1007946721546},
issn = {13875485},
journal = {Journal of VLSI Signal Processing Systems for Signal, Image, and Video Technology},
number = {2-3},
pages = {137--161},
title = {{A Hybrid Object-Oriented Very Low Bit Rate Video Codec}},
url = {https://link.springer.com/article/10.1023/A:1007946721546},
volume = {17},
year = {1997}
}

There are a large number of applications requiring the compression of video at Very Low Bit Rates (VLBR). Such applications include wireless video conferencing, video over the internet, multimedia database retrieval and remote sensing and monitoring. Recently, the MPEG-4 standardization effort has been a motivating factor to find a solution to this challenging problem. The existing approaches to this problem can generally be grouped into block-based, model-based, and object-oriented. Block-based approaches follow the traditional strategy of decoupling the image sequence into blocks, model-based approaches rely on complex 3-D models for specific objects that are encoded, and object-oriented approaches rely on analyzing the scene into differently moving objects. All three approaches exhibit potential problems. Block-based approaches tend to generate artifacts at the boundaries of the blocks, as well as to limit the minimum achievable bit-rate due to the fixed analysis structure of the scene. Model-based codecs are limited by the complex 3-D models of the objects to be encoded. On the other hand, object-oriented codecs can generate a significant overhead due to the analysis of the scene which needs to be transmitted, which in turn can be the limiting factor in achieving the target bit-rates. In this paper, we propose a hybrid object-oriented codec in which the correlations among the three information fields, e.g., motion, segmentation and intensity fields, are exploited both spatially and temporally. In the proposed method, additional intelligence is given to the decoder, resulting in a reduction of the required bandwidth. The residual information is analyzed into three different categories, i.e., occlusion, model failures, and global refinement. The residual information is encoded and transmitted across the channel with other side information. Experimental results are presented which demonstrate the effectiveness of the proposed approach.

@article{Guido1997d,
abstract = {In this paper we propose an algorithm for the optimal bit allocation among dependent quantizers for the minimum-maximum (MINMAX) distortion criterion. We compare this algorithm to the well-known Lagrange multiplier method for the minimum-average (MINAVE) distortion criterion. We point out the differences between these two distortion criteria, and their implications for coding applications. We argue that even though the MINAVE criterion is more popular, in many cases, the MINMAX criterion is more appropriate. We introduce the algorithms for solving the optimal bit allocation problem among dependent quantizers for both criteria and highlight the similarities and differences. We present the two algorithms using the same frame-work, which sheds new light on the relationship between the MINAVE and the MINMAX criteria. We point out that any problem which can be solved with the MINAVE criterion can also be solved with the MINMAX criterion, since both approaches are based on the same assumptions. {\textcopyright} 1997 Elsevier Science Ltd. All rights reserved.},
author = {Schuster, Guido M. and Katsaggelos, Aggelos K.},
doi = {10.1016/S0083-6656(97)00048-2},
issn = {00836656},
journal = {Vistas in Astronomy},
month = {jan},
number = {3},
pages = {427--437},
title = {{The minimum-average and minimum-maximum criteria in lossy compression}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0083665697000482},
volume = {41},
year = {1997}
}

In this paper we propose an algorithm for the optimal bit allocation among dependent quantizers for the minimum-maximum (MINMAX) distortion criterion. We compare this algorithm to the well-known Lagrange multiplier method for the minimum-average (MINAVE) distortion criterion. We point out the differences between these two distortion criteria, and their implications for coding applications. We argue that even though the MINAVE criterion is more popular, in many cases, the MINMAX criterion is more appropriate. We introduce the algorithms for solving the optimal bit allocation problem among dependent quantizers for both criteria and highlight the similarities and differences. We present the two algorithms using the same frame-work, which sheds new light on the relationship between the MINAVE and the MINMAX criteria. We point out that any problem which can be solved with the MINAVE criterion can also be solved with the MINMAX criterion, since both approaches are based on the same assumptions. © 1997 Elsevier Science Ltd. All rights reserved.

@article{MunGi1996,
abstract = {Restoration of image sequences is an important problem that can be encountered in many image processing applications, such as visual communications, robot guidance, and target tracking. The independent restoration of each frame in an image sequence is a suboptimal approach because the between-frame correlations are not explicitly taken into consideration. In this paper we address this problem by proposing a multichannel restoration approach. The multiple time-frames (channels) of the image sequence are restored simultaneously by using a multichannel regularized least-squares formulation of the problem. The regularization operator captures both within- and between-frame (channel) properties of the image sequence with the explicit use of the displacement vector field. We propose a number of different approaches to obtain the multichannel regularization operator, as well as an algorithm to iteratively compute the restored images. We present experiments that demonstrate the value of the proposed multichannel approach. {\textcopyright} 1996 Academic Press, Inc.},
author = {Choi, Mun Gi and Galatsanos, Nikolas P. and Katsaggelos, Aggelos K.},
doi = {10.1006/jvci.1996.0022},
issn = {10473203},
journal = {Journal of Visual Communication and Image Representation},
month = {sep},
number = {3},
pages = {244--258},
title = {{Multichannel Regularized Iterative Restoration of Motion Compensated Image Sequences}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S104732039690022X},
volume = {7},
year = {1996}
}

Restoration of image sequences is an important problem that can be encountered in many image processing applications, such as visual communications, robot guidance, and target tracking. The independent restoration of each frame in an image sequence is a suboptimal approach because the between-frame correlations are not explicitly taken into consideration. In this paper we address this problem by proposing a multichannel restoration approach. The multiple time-frames (channels) of the image sequence are restored simultaneously by using a multichannel regularized least-squares formulation of the problem. The regularization operator captures both within- and between-frame (channel) properties of the image sequence with the explicit use of the displacement vector field. We propose a number of different approaches to obtain the multichannel regularization operator, as well as an algorithm to iteratively compute the restored images. We present experiments that demonstrate the value of the proposed multichannel approach. © 1996 Academic Press, Inc.

@article{tull1996iterative,
author = {Tull, Damon L. and Katsaggelos, Aggelos K.},
doi = {10.1117/1.601108},
issn = {0091-3286},
journal = {Optical Engineering},
month = {dec},
number = {12},
pages = {3460},
publisher = {SPIE},
title = {{Iterative restoration of fast‐moving objects in dynamic image sequences}},
url = {http://opticalengineering.spiedigitallibrary.org/article.aspx?doi=10.1117/1.601108},
volume = {35},
year = {1996}
}

@article{molina1996compound,
abstract = {Over the last few years, a growing number of researchers from varied disciplines have been utilizing Markov random fields (MRF) models for developing optimal, robust algorithms for various problems, such as texture analysis, image synthesis, image restoration, classification and segmentation, surface reconstruction, integration of several low level vision modules and sensor fusion. While linear-shift invariant (LSI) models have been generally used for image restoration in astronomy, no much work has been reported on the use of more complex models in this area. In this paper we examine the use of Compound Gaussian Markov Random Fields, (CGMRF), a non LSI model that preserves image discontinuities, to restore astronomical images. Problems on the application of the model arising from the high dynamic range and severe blurring of astronomical images are addressed and two new methods to estimate the real underlying image are proposed. The methods are tested on real astronomical images. Copyright {\textcopyright} 1996 Elsevier Science Ltd.},
author = {Molina, R. and Katsaggelos, A.K. and Mateos, J. and Abad, J.},
doi = {10.1016/S0083-6656(96)00039-6},
issn = {00836656},
journal = {Vistas in Astronomy},
month = {jan},
number = {4},
pages = {539--546},
publisher = {North-Holland},
title = {{Compound Gauss-Markov random fields for astronomical image restoration}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0083665696000396},
volume = {40},
year = {1996}
}

Over the last few years, a growing number of researchers from varied disciplines have been utilizing Markov random fields (MRF) models for developing optimal, robust algorithms for various problems, such as texture analysis, image synthesis, image restoration, classification and segmentation, surface reconstruction, integration of several low level vision modules and sensor fusion. While linear-shift invariant (LSI) models have been generally used for image restoration in astronomy, no much work has been reported on the use of more complex models in this area. In this paper we examine the use of Compound Gaussian Markov Random Fields, (CGMRF), a non LSI model that preserves image discontinuities, to restore astronomical images. Problems on the application of the model arising from the high dynamic range and severe blurring of astronomical images are addressed and two new methods to estimate the real underlying image are proposed. The methods are tested on real astronomical images. Copyright © 1996 Elsevier Science Ltd.

@article{tom1996multichannel,
abstract = {Previous work has demonstrated the effectiveness of the expectation-maximization algorithm to restore noisy and blurred single- channel images and simultaneously identify its blur. In addition, a gen- eral framework for processing multichannel images using single-channel techniques has been developed. The authors combine and extend the two approaches to the simultaneous blur identification and restoration of multichannel images. Explicit equations for that purpose are developed for the general case when cross-channel degradations are present. An important difference from the single-channel problem is that the cross power spectra are complex quantities, which further complicates the analysis of the algorithm. The proposed algorithm is very effective at restoring multichannel images, as is demonstrated experimentally.},
author = {Tom, Brian C. and Lay, Kuen-Tsair and Katsaggelos, Aggelos K.},
doi = {10.1117/1.600876},
issn = {0091-3286},
journal = {Optical Engineering},
month = {jan},
number = {1},
pages = {241},
publisher = {SPIE},
title = {{Multichannel image identification and restoration using the expectation‐maximization algorithm}},
url = {http://opticalengineering.spiedigitallibrary.org/article.aspx?doi=10.1117/1.600876},
volume = {35},
year = {1996}
}

Previous work has demonstrated the effectiveness of the expectation-maximization algorithm to restore noisy and blurred single- channel images and simultaneously identify its blur. In addition, a gen- eral framework for processing multichannel images using single-channel techniques has been developed. The authors combine and extend the two approaches to the simultaneous blur identification and restoration of multichannel images. Explicit equations for that purpose are developed for the general case when cross-channel degradations are present. An important difference from the single-channel problem is that the cross power spectra are complex quantities, which further complicates the analysis of the algorithm. The proposed algorithm is very effective at restoring multichannel images, as is demonstrated experimentally.

@article{banham1996spatially,
abstract = {In this paper, we present a new spatially adaptive approach to the restoration of noisy blurred images, which is particularly effective at producing sharp deconvolution while suppressing the noise in the flat regions of an image. This is accomplished through a multiscale Kalman smoothing filter applied to a prefiltered observed image in the discrete, separable, 2-D wavelet domain. The prefiltering step involves constrained least-squares filtering based on optimal choices for the regularization parameter. This leads to a reduction in the support of the required state vectors of the multiscale restoration filter in the wavelet domain and improvement in the computational efficiency of the multiscale filter. The proposed method has the benefit that the majority of the regularization, or noise suppression, of the restoration is accomplished by the efficient multiscale filtering of wavelet detail coefficients ordered on quadtrees. Not only does this lead to potential parallel implementation schemes, but it permits adaptivity to the local edge information in the image. In particular, this method changes filter parameters depending on scale, local signal-to-noise ratio (SNR), and orientation. Because the wavelet detail coefficients are a manifestation of the multiscale edge information in an image, this algorithm may be viewed as an "edge-adaptive" multiscale restoration approach. {\textcopyright} 1996 IEEE.},
author = {Banham, M.R. and Katsaggelos, A.K.},
doi = {10.1109/83.491338},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
month = {apr},
number = {4},
pages = {619--634},
publisher = {IEEE},
title = {{Spatially adaptive wavelet-based multiscale image restoration}},
url = {http://ieeexplore.ieee.org/document/491338/},
volume = {5},
year = {1996}
}

In this paper, we present a new spatially adaptive approach to the restoration of noisy blurred images, which is particularly effective at producing sharp deconvolution while suppressing the noise in the flat regions of an image. This is accomplished through a multiscale Kalman smoothing filter applied to a prefiltered observed image in the discrete, separable, 2-D wavelet domain. The prefiltering step involves constrained least-squares filtering based on optimal choices for the regularization parameter. This leads to a reduction in the support of the required state vectors of the multiscale restoration filter in the wavelet domain and improvement in the computational efficiency of the multiscale filter. The proposed method has the benefit that the majority of the regularization, or noise suppression, of the restoration is accomplished by the efficient multiscale filtering of wavelet detail coefficients ordered on quadtrees. Not only does this lead to potential parallel implementation schemes, but it permits adaptivity to the local edge information in the image. In particular, this method changes filter parameters depending on scale, local signal-to-noise ratio (SNR), and orientation. Because the wavelet detail coefficients are a manifestation of the multiscale edge information in an image, this algorithm may be viewed as an "edge-adaptive" multiscale restoration approach. © 1996 IEEE.

@article{chah1995linear,
author = {Chah, C.L. and Katsaggelos, A.K. and Sahakian, A.V.},
doi = {10.1109/83.413179},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
number = {9},
pages = {1328--1333},
publisher = {IEEE},
title = {{Linear-quadratic noise-smoothing filters for quantum-limited images}},
url = {http://ieeexplore.ieee.org/document/413179/},
volume = {4},
year = {1995}
}

@article{zhu1995regularized,
abstract = {Multichannel images are the multiple image planes (channels) obtained by imaging the same scene using multiple sensors. The validity of multichannel restoration where both within- and between-channel relations are incorporated has already been established using both stochastic and deterministic restoration filters. However, it has been demonstrated that stochastic multichannel filters are extremely sensitive to the estimates of the between-channel statistics. In this paper we avoid the problems associated with multichannel stochastic filters by proposing deterministic multichannel filters that do not require any prior knowledge about either the statistics of the multichannel image or the noise. Regularization based on the multichannel cross-validation function is used to obtain these filters. We examine their relation to multichannel linear minimum mean square error restoration filters and we propose a technique to estimate the variance of the noise. Finally, we show experiments where we test the proposed filters and noise variance estimator using color images. {\textcopyright} 1994 Academic Press. All rights reserved.},
author = {Zhu, W.W. and Galatsanos, N.P. and Katsaggelos, A.K.},
doi = {10.1006/gmip.1995.1005},
institution = {SPIE INTERNATIONAL SOCIETY FOR OPTICAL},
issn = {10773169},
journal = {Graphical Models and Image Processing},
month = {jan},
number = {1},
pages = {38--54},
publisher = {Academic Press},
title = {{Regularized Multichannel Restoration Using Cross-Validation}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1077316985710052},
volume = {57},
year = {1995}
}

Multichannel images are the multiple image planes (channels) obtained by imaging the same scene using multiple sensors. The validity of multichannel restoration where both within- and between-channel relations are incorporated has already been established using both stochastic and deterministic restoration filters. However, it has been demonstrated that stochastic multichannel filters are extremely sensitive to the estimates of the between-channel statistics. In this paper we avoid the problems associated with multichannel stochastic filters by proposing deterministic multichannel filters that do not require any prior knowledge about either the statistics of the multichannel image or the noise. Regularization based on the multichannel cross-validation function is used to obtain these filters. We examine their relation to multichannel linear minimum mean square error restoration filters and we propose a technique to estimate the variance of the noise. Finally, we show experiments where we test the proposed filters and noise variance estimator using color images. © 1994 Academic Press. All rights reserved.

@article{brailean1995noise,
abstract = {In this paper, a thorough review is presented of noise reduction filters for digital image sequences. Detailed descriptions of several spatiotemporal and temporal noise reduction algorithms are provided. To aid in comparing between these different algorithms, we classify them based on their support (i.e., 3-D or 1-D filter) and whether or not motion compensation is employed. Several algorithms from each of the four categories are implemented and tested on real sequences degraded to various signal-to-noise ratios. These experimental results are discussed and analyzed to determine the overall advantages and disadvantages of the four general classifications, as well as, the individual filters. {\textcopyright} 1995 IEEE},
author = {Brailean, J.C. and Kleihorst, R.P. and Efstratiadis, Serafim and Katsaggelos, A.K. and Lagendijk, R.L.},
doi = {10.1109/5.406412},
issn = {00189219},
journal = {Proceedings of the IEEE},
number = {9},
pages = {1272--1292},
publisher = {IEEE},
title = {{Noise reduction filters for dynamic image sequences: a review}},
url = {http://ieeexplore.ieee.org/document/406412/},
volume = {83},
year = {1995}
}

In this paper, a thorough review is presented of noise reduction filters for digital image sequences. Detailed descriptions of several spatiotemporal and temporal noise reduction algorithms are provided. To aid in comparing between these different algorithms, we classify them based on their support (i.e., 3-D or 1-D filter) and whether or not motion compensation is employed. Several algorithms from each of the four categories are implemented and tested on real sequences degraded to various signal-to-noise ratios. These experimental results are discussed and analyzed to determine the overall advantages and disadvantages of the four general classifications, as well as, the individual filters. © 1995 IEEE

@article{Michael1995,
abstract = {This paper considers the concept of robust esti- mation in regularized image restoration. Robust functionals are employed for the representation of both the noise and the signal statistics. Such functionals allow the efficient suppression of a wide variety of noise processes and permit the reconstruction of sharper edges than their quadratic counterparts. A new class of robust entropic functionals is introduced, which operates only on the high-frequency content of the signal and reflects sharp deviations in the signal distribution. This class of functionals can also incorporate prior structural information regarding the original image, in a way similar to the maximum information principle. The convergence properties of robust iterative algorithms are studied for continuously and noncontinuously differentiable functionals. The definition of the robust approach is completed by introducing a method for the optimal selection of the regularization parameter. This method utilizes the structure of robust estimators that lack analytic specification. The properties of robust algorithms are demonstrated through restoration examples in different noise environments. {\textcopyright} 1995 IEEE.},
author = {Zervakis, M.E. and Katsaggelos, A.K. and Kwon, T.M.},
doi = {10.1109/83.388078},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
month = {jun},
number = {6},
pages = {752--773},
title = {{A class of robust entropic functionals for image restoration}},
url = {http://ieeexplore.ieee.org/document/388078/},
volume = {4},
year = {1995}
}

This paper considers the concept of robust esti- mation in regularized image restoration. Robust functionals are employed for the representation of both the noise and the signal statistics. Such functionals allow the efficient suppression of a wide variety of noise processes and permit the reconstruction of sharper edges than their quadratic counterparts. A new class of robust entropic functionals is introduced, which operates only on the high-frequency content of the signal and reflects sharp deviations in the signal distribution. This class of functionals can also incorporate prior structural information regarding the original image, in a way similar to the maximum information principle. The convergence properties of robust iterative algorithms are studied for continuously and noncontinuously differentiable functionals. The definition of the robust approach is completed by introducing a method for the optimal selection of the regularization parameter. This method utilizes the structure of robust estimators that lack analytic specification. The properties of robust algorithms are demonstrated through restoration examples in different noise environments. © 1995 IEEE.

@article{James1995,
abstract = {In this paper, we develop a recursive model-based maximum a posteriori (MAP) estimator that simultaneously estimates the displacement vector field (DVF) and the intensity field from a noisy-blurred image sequence. Current motion-compensated spatio-temporal noise filters treat the estimation of the DVF as a preprocessing step. Generally, no attempt is made to verify the accuracy of these estimates prior to their use in the filter. By simultaneously estimating these two fields, we establish a link between the two estimators. It is through this link that the DVF estimate and its corresponding accuracy information are shared with the other intensity estimator, and vice versa. To model the DVF and the intensity field, we use coupled Gauss-Markov (CGM) models. A CGM model consists of two levels: An upper level, which is made up of several submodels with various characteristics, and a lower level or line field, which governs the transitions between the submodels. The CGM models are well suited for estimating the displacement and intensity fields since the resulting estimates preserve the boundaries between the stationary areas present in both fields. Detailed line fields are proposed for the modeling of these boundaries, which also take into account the correlations that exist between these two fields. A Kalman-type estimator results, followed by a decision criterion for choosing the appropriate set of line fields. Several experiments using noisy and noisy-blurred image sequences demonstrate the superior performance of the proposed algorithm with respect to prediction error and mean-square error. {\textcopyright} 1995 IEEE.},
author = {Brailean, J.C. and Katsaggelos, A.K.},
doi = {10.1109/83.413168},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
number = {9},
pages = {1236--1251},
title = {{Simultaneous recursive displacement estimation and restoration of noisy-blurred image sequences}},
url = {http://ieeexplore.ieee.org/document/413168/},
volume = {4},
year = {1995}
}

In this paper, we develop a recursive model-based maximum a posteriori (MAP) estimator that simultaneously estimates the displacement vector field (DVF) and the intensity field from a noisy-blurred image sequence. Current motion-compensated spatio-temporal noise filters treat the estimation of the DVF as a preprocessing step. Generally, no attempt is made to verify the accuracy of these estimates prior to their use in the filter. By simultaneously estimating these two fields, we establish a link between the two estimators. It is through this link that the DVF estimate and its corresponding accuracy information are shared with the other intensity estimator, and vice versa. To model the DVF and the intensity field, we use coupled Gauss-Markov (CGM) models. A CGM model consists of two levels: An upper level, which is made up of several submodels with various characteristics, and a lower level or line field, which governs the transitions between the submodels. The CGM models are well suited for estimating the displacement and intensity fields since the resulting estimates preserve the boundaries between the stationary areas present in both fields. Detailed line fields are proposed for the modeling of these boundaries, which also take into account the correlations that exist between these two fields. A Kalman-type estimator results, followed by a decision criterion for choosing the appropriate set of line fields. Several experiments using noisy and noisy-blurred image sequences demonstrate the superior performance of the proposed algorithm with respect to prediction error and mean-square error. © 1995 IEEE.

@article{kang1995general,
abstract = {The determination of the regularization parameter is an important issue in regularized image restoration, since it controls the trade-off between fidelity to the data and smoothness of the solution. A number of approaches have been developed in determining this parameter. In this paper, a new paradigm is adopted, according to which the required prior information is extracted from the available data at the previous iteration step, i.e., the partially restored image at each step. We propose the use of a regularization functional instead of a constant regularization parameter. The properties such a regularization functional should satisfy are investigated, and two specific forms of it are proposed. An iterative algorithm is proposed for obtaining a restored image. The regularization functional is defined in terms of the restored image at each iteration step, therefore allowing for the simultaneous determination of its value and the restoration of the degraded image. Both proposed iteration adaptive regularization functionals are shown to result in a smoothing functional with a global minimum, so that its iterative optimization does not depend on the initial conditions. The convergence of the algorithm is established and experimental results are shown. {\textcopyright} 1995 IEEE.},
author = {Katsaggelos, Aggelos K. A.K. and {Moon Gi Kang} and Katsaggelos, Aggelos K. A.K.},
doi = {10.1109/83.382494},
issn = {19410042},
journal = {IEEE Transactions on Image Processing},
month = {may},
number = {5},
pages = {594--602},
publisher = {IEEE},
title = {{General choice of the regularization functional in regularized image restoration}},
url = {http://ieeexplore.ieee.org/document/382494/},
volume = {4},
year = {1995}
}

The determination of the regularization parameter is an important issue in regularized image restoration, since it controls the trade-off between fidelity to the data and smoothness of the solution. A number of approaches have been developed in determining this parameter. In this paper, a new paradigm is adopted, according to which the required prior information is extracted from the available data at the previous iteration step, i.e., the partially restored image at each step. We propose the use of a regularization functional instead of a constant regularization parameter. The properties such a regularization functional should satisfy are investigated, and two specific forms of it are proposed. An iterative algorithm is proposed for obtaining a restored image. The regularization functional is defined in terms of the restored image at each iteration step, therefore allowing for the simultaneous determination of its value and the restoration of the degraded image. Both proposed iteration adaptive regularization functionals are shown to result in a smoothing functional with a global minimum, so that its iterative optimization does not depend on the initial conditions. The convergence of the algorithm is established and experimental results are shown. © 1995 IEEE.

@article{brailean1995recursive,
abstract = {In this paper, a recursive model-based algorithm for obtaining the maximum a posteriori (MAP) estimate of the displacement vector field (DVF) from successive image frames of an image sequence is presented. To model the DVF, we develop a nonstationary vector field model called the vector coupled Gauss-Markov (VCGM) model. The VCGM model consists of two levels: an upper level, which is made up of several submodels with various characteristics, and a lower level or line process, which governs the transitions between the submodels. A detailed line process is proposed. The VCGM model is well suited for estimating the DVF since the resulting estimates preserve the boundaries between the differently moving areas in an image sequence. A Kalman type estimator results, followed by a decision criterion for choosing the appropriate line process. Several experiments demonstrate the superior performance of the proposed algorithm with respect to prediction error, interpolation error, and robustness to noise. {\textcopyright} 1995 IEEE},
author = {Brailean, J.C. and Katsaggelos, A.K.},
doi = {10.1109/83.370672},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
month = {apr},
number = {4},
pages = {416--429},
publisher = {IEEE},
title = {{A recursive nonstationary MAP displacement vector field estimation algorithm}},
url = {http://ieeexplore.ieee.org/document/370672/},
volume = {4},
year = {1995}
}

In this paper, a recursive model-based algorithm for obtaining the maximum a posteriori (MAP) estimate of the displacement vector field (DVF) from successive image frames of an image sequence is presented. To model the DVF, we develop a nonstationary vector field model called the vector coupled Gauss-Markov (VCGM) model. The VCGM model consists of two levels: an upper level, which is made up of several submodels with various characteristics, and a lower level or line process, which governs the transitions between the submodels. A detailed line process is proposed. The VCGM model is well suited for estimating the DVF since the resulting estimates preserve the boundaries between the differently moving areas in an image sequence. A Kalman type estimator results, followed by a decision criterion for choosing the appropriate line process. Several experiments demonstrate the superior performance of the proposed algorithm with respect to prediction error, interpolation error, and robustness to noise. © 1995 IEEE

@article{chan1993maximum,
abstract = {In this paper, we develop an algorithm for obtain- ing the maximum likelihood (ML) estimate of the displacement vector field (DVF) from two consecutive image frames of an image sequence acquired under quantum-limited conditions. The estimation of the DVF has applications in temporal filtering, object tracking, stereo matching, and frame registration in low-light level image sequences as well as low-dose clinical x-ray image sequences. In the latter case, a controlled x-ray dosage reduction may be utilized to lower the radiation exposure to the patient and the medical staff. The quantum-limited effect is modeled as an undesirable, Poisson-distributed, signal-dependent noise artifact. A Fisher-Bayesian formulation is used in this paper to estimate the DVF and a block component search algorithm is employed in obtaining the solution. Several experiments involving a phantom sequence and a teleconferencing image sequence with realistic motion demonstrate the effectiveness of this estimator in obtaining the DVF under severe quantum noise conditions (20-25 events/pixel). {\textcopyright} 1995 IEEE.},
author = {Chan, C.L. and Katsaggelos, A.K.},
doi = {10.1109/83.388077},
institution = {SPIE},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
month = {jun},
number = {6},
pages = {743--751},
title = {{Iterative maximum likelihood displacement field estimation in quantum-limited image sequences}},
url = {http://ieeexplore.ieee.org/document/388077/},
volume = {4},
year = {1995}
}

In this paper, we develop an algorithm for obtain- ing the maximum likelihood (ML) estimate of the displacement vector field (DVF) from two consecutive image frames of an image sequence acquired under quantum-limited conditions. The estimation of the DVF has applications in temporal filtering, object tracking, stereo matching, and frame registration in low-light level image sequences as well as low-dose clinical x-ray image sequences. In the latter case, a controlled x-ray dosage reduction may be utilized to lower the radiation exposure to the patient and the medical staff. The quantum-limited effect is modeled as an undesirable, Poisson-distributed, signal-dependent noise artifact. A Fisher-Bayesian formulation is used in this paper to estimate the DVF and a block component search algorithm is employed in obtaining the solution. Several experiments involving a phantom sequence and a teleconferencing image sequence with realistic motion demonstrate the effectiveness of this estimator in obtaining the DVF under severe quantum noise conditions (20-25 events/pixel). © 1995 IEEE.

@article{yang1995projection,
abstract = {At the present time, block-transform coding is probably the most popular approach for image compression. For this approach, the compressed images are decoded using only the transmitted transform data. In this paper, we formulate image decoding as an image recovery problem. According to this approach, the decoded image is reconstructed using not only the transmitted data but, in addition, the prior knowledge that images before compression do not display between-block discontinuities. A spatially adaptive image recovery algorithm is proposed based on the theory of projections onto convex sets. Apart from the data constraint set, this algorithm uses another new constraint set that enforces between-block smoothness. The novelty of this set is that it captures both the local statistical properties of the image and the human perceptual characteristics. A simplified spatially adaptive recovery algorithm is also proposed, and the analysis of its computational complexity is presented. Numerical experiments are shown that demonstrate that the proposed algorithms work better than both the JPEG deblocking recommendation and our previous projection-based image decoding approach. {\textcopyright} 1995 IEEE},
author = {{Yongyi Yang} and Galatsanos, Nikolas P. N.P. and Katsaggelos, Aggelos K. A.K. and Yang, Yongyi and Galatsanos, Nikolas P. N.P. and Katsaggelos, Aggelos K. A.K.},
doi = {10.1109/83.392332},
institution = {SPIE INTERNATIONAL SOCIETY FOR OPTICAL},
issn = {19410042},
journal = {IEEE Transactions on Image Processing},
month = {jul},
number = {7},
pages = {896--908},
publisher = {IEEE},
title = {{Projection-Based Spatially Adaptive Reconstruction of Block-Transform Compressed Images}},
url = {http://ieeexplore.ieee.org/document/392332/},
volume = {4},
year = {1995}
}

At the present time, block-transform coding is probably the most popular approach for image compression. For this approach, the compressed images are decoded using only the transmitted transform data. In this paper, we formulate image decoding as an image recovery problem. According to this approach, the decoded image is reconstructed using not only the transmitted data but, in addition, the prior knowledge that images before compression do not display between-block discontinuities. A spatially adaptive image recovery algorithm is proposed based on the theory of projections onto convex sets. Apart from the data constraint set, this algorithm uses another new constraint set that enforces between-block smoothness. The novelty of this set is that it captures both the local statistical properties of the image and the human perceptual characteristics. A simplified spatially adaptive recovery algorithm is also proposed, and the analysis of its computational complexity is presented. Numerical experiments are shown that demonstrate that the proposed algorithms work better than both the JPEG deblocking recommendation and our previous projection-based image decoding approach. © 1995 IEEE

@article{katsaggelos1995spatially,
abstract = {This article develops an iterative spatially adaptive regularized image restoration algorithm. The proposed algorithm is based on the minimization of a weighted smoothing functional. The weighting matrices are defined as functions of the partially restored image at each iteration step. As a result, no prior knowledge about the image and the noise is required, but the weighting matrices as well as the regularization parameter are updated based on the restored image at every step. Conditions for the convexity of the weighted smoothing functional and for the convergence of the iterative algorithm are established for a unique global solution which does not depend on initial conditions. Experimental results are shown with astronomical images which demonstrate the effectiveness of the proposed algorithm. Copyright {\textcopyright} 1995 Wiley Periodicals, Inc., A Wiley Company},
author = {Katsaggelos, Aggelos K. and Kang, Moon Gi},
doi = {10.1002/ima.1850060404},
issn = {08999457},
journal = {International Journal of Imaging Systems and Technology},
number = {4},
pages = {305--313},
publisher = {John Wiley \& Sons, Inc. New York},
title = {{Spatially adaptive iterative algorithm for the restoration of astronomical images}},
url = {https://onlinelibrary.wiley.com/doi/10.1002/ima.1850060404},
volume = {6},
year = {1995}
}

This article develops an iterative spatially adaptive regularized image restoration algorithm. The proposed algorithm is based on the minimization of a weighted smoothing functional. The weighting matrices are defined as functions of the partially restored image at each iteration step. As a result, no prior knowledge about the image and the noise is required, but the weighting matrices as well as the regularization parameter are updated based on the restored image at every step. Conditions for the convexity of the weighted smoothing functional and for the convergence of the iterative algorithm are established for a unique global solution which does not depend on initial conditions. Experimental results are shown with astronomical images which demonstrate the effectiveness of the proposed algorithm. Copyright © 1995 Wiley Periodicals, Inc., A Wiley Company

@article{banham1994low,
abstract = {In this paper, we present a novel coding technique that makes use of the nonstationary characteristics of an image sequence displacement field to estimate and encode motion information. We utilize an MPEG style codec in which the anchor frames in a sequence are encoded with a hybrid approach using quadtree, DCT, and wavelet-based coding techniques. A quadtree structured approach is also utilized for the interframe information. The main objective of the overall design is to demonstrate the coding potential of a newly developed motion estimator called the coupled linearized MAP (CLMAP) estimator. This estimator can be used as a means for producing motion vectors that may be regenerated at the decoder with a coarsely quantized error term created in the encoder. The motion estimator generates highly accurate motion estimates from this coarsely quantized data. This permits the elimination of a separately coded displaced frame difference (DFD) and coded motion vectors. For low bit rate applications, this is especially important because the overhead associated with the transmission of motion vectors may become prohibitive. We exploit both the advantages of the nonstationary motion estimator and the effective compression of the anchor frame coder to improve the visual quality of reconstructed QCIF format color image sequences at low bit rates. Comparisons are made with other video coding methods, including the H.261 and MPEG standards and a pel-recursive-based codec. {\textcopyright} 1994 IEEE},
author = {Banham, M.R. and Brailean, J.C. and Chan, C.L. and Katsaggelos, A.K.},
doi = {10.1109/83.334979},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
number = {5},
pages = {652--665},
publisher = {IEEE},
title = {{Low bit rate video coding using robust motion vector regeneration in the decoder}},
url = {http://ieeexplore.ieee.org/document/334979/},
volume = {3},
year = {1994}
}

In this paper, we present a novel coding technique that makes use of the nonstationary characteristics of an image sequence displacement field to estimate and encode motion information. We utilize an MPEG style codec in which the anchor frames in a sequence are encoded with a hybrid approach using quadtree, DCT, and wavelet-based coding techniques. A quadtree structured approach is also utilized for the interframe information. The main objective of the overall design is to demonstrate the coding potential of a newly developed motion estimator called the coupled linearized MAP (CLMAP) estimator. This estimator can be used as a means for producing motion vectors that may be regenerated at the decoder with a coarsely quantized error term created in the encoder. The motion estimator generates highly accurate motion estimates from this coarsely quantized data. This permits the elimination of a separately coded displaced frame difference (DFD) and coded motion vectors. For low bit rate applications, this is especially important because the overhead associated with the transmission of motion vectors may become prohibitive. We exploit both the advantages of the nonstationary motion estimator and the effective compression of the anchor frame coder to improve the visual quality of reconstructed QCIF format color image sequences at low bit rates. Comparisons are made with other video coding methods, including the H.261 and MPEG standards and a pel-recursive-based codec. © 1994 IEEE

@article{Mark1994,
abstract = {In this paper, we present a new matrix vector formulation of a wavelet-based subband decomposition. This formulation allows for the decomposition of both the convolution operator and the signal in the subband domain. With this approach, any single channel linear space-invariant filtering problem can be cast into a multichannel framework. We apply this decomposition to the linear space-invariant image restoration problem and propose a family of multichannel linear minimum mean square error (LMMSE) restoration ffiters. These filters explicitly incorporate both within and between subband (channel) relations of the decomposed image. Since only within channel stationarity is assumed in the image model, this approach presents a new method for modeling the nonstationarity of images. Experimental results are presented which test the proposed multichannel LMMSE filters. These experiments show that if accurate estimates of the subband statistics are available, the proposed multichannel filters provide major improvements over the traditional single channel filters. {\textcopyright} 1994 IEEE},
author = {Banham, M.R. and Galatsanos, N.P. and Gonzalez, H.L. and Katsaggelos, A.K.},
doi = {10.1109/83.336250},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
number = {6},
pages = {821--833},
publisher = {IEEE},
title = {{Multichannel restoration of single channel images using a wavelet-based subband decomposition}},
url = {http://ieeexplore.ieee.org/document/336250/},
volume = {3},
year = {1994}
}

In this paper, we present a new matrix vector formulation of a wavelet-based subband decomposition. This formulation allows for the decomposition of both the convolution operator and the signal in the subband domain. With this approach, any single channel linear space-invariant filtering problem can be cast into a multichannel framework. We apply this decomposition to the linear space-invariant image restoration problem and propose a family of multichannel linear minimum mean square error (LMMSE) restoration ffiters. These filters explicitly incorporate both within and between subband (channel) relations of the decomposed image. Since only within channel stationarity is assumed in the image model, this approach presents a new method for modeling the nonstationarity of images. Experimental results are presented which test the proposed multichannel LMMSE filters. These experiments show that if accurate estimates of the subband statistics are available, the proposed multichannel filters provide major improvements over the traditional single channel filters. © 1994 IEEE

@article{tom1994motion,
abstract = {In this paper an approach for estimating the motion of arteries in digital angiographic image sequences is proposed. Binary skeleton images are registered using an elastic registration algorithm in order to estimate the motion of the corresponding arteries. This algorithm operates recursively on the skeleton images by considering an autoregressive (AR) model of the deformation in conjunction with a dynamic programming (DP) algorithm. The AR model is used at the pixel level and provides a suitable cost function to DP through the innovation process. In addition, a moving average (MA) model for the motion of the entire skeleton is used in combination with the local AR model for improved registration results. The performance of this motion estimation method is demonstrated on simulated and real digital angiographic image sequences. It is shown that motion estimation using elastic registration of skeletons is very successful especially with low contrast and noisy angiographic images. {\textcopyright} 1994 IEEE},
author = {Tom, B.C.S. and Efstratiadis, S.N. and Katsaggelos, A.K.},
doi = {10.1109/42.310876},
issn = {02780062},
journal = {IEEE Transactions on Medical Imaging},
number = {3},
pages = {450--460},
publisher = {IEEE},
title = {{Motion estimation of skeletonized angiographic images using elastic registration}},
url = {http://ieeexplore.ieee.org/document/310876/},
volume = {13},
year = {1994}
}

In this paper an approach for estimating the motion of arteries in digital angiographic image sequences is proposed. Binary skeleton images are registered using an elastic registration algorithm in order to estimate the motion of the corresponding arteries. This algorithm operates recursively on the skeleton images by considering an autoregressive (AR) model of the deformation in conjunction with a dynamic programming (DP) algorithm. The AR model is used at the pixel level and provides a suitable cost function to DP through the innovation process. In addition, a moving average (MA) model for the motion of the entire skeleton is used in combination with the local AR model for improved registration results. The performance of this motion estimation method is demonstrated on simulated and real digital angiographic image sequences. It is shown that motion estimation using elastic registration of skeletons is very successful especially with low contrast and noisy angiographic images. © 1994 IEEE

@article{goyette1994improving,
abstract = {In this article the point spread function of autoradiography is experimentally measured and used with digital image restoration techniques to improve the resolution of autoradiographs. We compare two regularized iterative image restoration algorithms applied to autoradiography. A nonlinear filter is used for the removal of film grain noise prior to restoration. Our results indicate a 27 percent improvement in resolution.},
author = {Goyette, J. A. and Lapin, G. D. and Kang, M. G. and Katsaggelos, A. K.},
issn = {07395175},
journal = {IEEE Engineering in Medicine and Biology Magazine},
number = {4},
pages = {571--574},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
title = {{Improving autoradiograph resolution using image restoration techniques}},
volume = {13},
year = {1994}
}

In this article the point spread function of autoradiography is experimentally measured and used with digital image restoration techniques to improve the resolution of autoradiographs. We compare two regularized iterative image restoration algorithms applied to autoradiography. A nonlinear filter is used for the removal of film grain noise prior to restoration. Our results indicate a 27 percent improvement in resolution.

@article{kang1994frequency,
author = {Kang, Moon Gi and Katsaggelos, Aggelos K.},
doi = {10.1117/12.181245},
issn = {0091-3286},
journal = {Optical Engineering},
month = {oct},
number = {10},
pages = {3222},
publisher = {SPIE},
title = {{Frequency-domain adaptive iterative image restoration and evaluation of the regularization parameter}},
url = {http://opticalengineering.spiedigitallibrary.org/article.aspx?doi=10.1117/12.181245},
volume = {33},
year = {1994}
}

@article{Michael1993,
abstract = {The pel-recursive approach to motion estimation has been widely studied for compensating progressively scanned, moderate-resolution video. Although pel-recursive algorithms may not be suitable for application to interlaced high-definition television (HDTV), the underlying principle of backward motion compensation, upon which pel-recursive algorithms are based, can be exploited to improve the existing motion compensation algorithms. This paper proposes applying a backward approach to motion compensation to improve the performance of standard block-based algorithms for motion-compensated interlaced HDTV sequences. First, we describe a framework for motion compensation in which motion information is parameterized by a motion operator and a domain for that operator. Within this framework, we characterize the type of motion information represented by forward (e.g. block-based) and backward approaches to motion compensation. We propose a method for combining these two sources of motion information to form an optimal motion-compensated prediction. Simulations on two interlaced HDTV sequences demonstrate performance improvements between 1 and 2 dB over standard block-based methods. {\textcopyright} 1993.},
author = {Orchard, Michael T. and Katsaggelos, Aggelos K.},
doi = {10.1016/0923-5965(93)90012-I},
issn = {09235965},
journal = {Signal Processing: Image Communication},
keywords = {Motion compensation,interlaced HDTV,pel-recursive motion estimation},
month = {dec},
number = {5-6},
pages = {487--501},
title = {{Backward motion compensation for interlaced HDTV}},
url = {https://linkinghub.elsevier.com/retrieve/pii/092359659390012I},
volume = {5},
year = {1993}
}

The pel-recursive approach to motion estimation has been widely studied for compensating progressively scanned, moderate-resolution video. Although pel-recursive algorithms may not be suitable for application to interlaced high-definition television (HDTV), the underlying principle of backward motion compensation, upon which pel-recursive algorithms are based, can be exploited to improve the existing motion compensation algorithms. This paper proposes applying a backward approach to motion compensation to improve the performance of standard block-based algorithms for motion-compensated interlaced HDTV sequences. First, we describe a framework for motion compensation in which motion information is parameterized by a motion operator and a domain for that operator. Within this framework, we characterize the type of motion information represented by forward (e.g. block-based) and backward approaches to motion compensation. We propose a method for combining these two sources of motion information to form an optimal motion-compensated prediction. Simulations on two interlaced HDTV sequences demonstrate performance improvements between 1 and 2 dB over standard block-based methods. © 1993.

@article{katsaggelos1993two,
author = {Katsaggelos, A K},
journal = {Newsl. STScI's Image Restoration Proj},
pages = {30--37},
title = {{Two possible approaches to restoration of HST images.}},
volume = {1},
year = {1993}
}

@article{chan1993image,
abstract = {Clinical angiography—the procedure of acquiring radiographic (fluoroscopic) image sequences of patients from x-ray based medical systems—has unquestionably aided cardiologists in their assessment of coronary disease. During such trials, however, literally hundreds of x-ray images are gathered, thereby putting these patients and particularly the medical staff at risk. It is desirable to lower the clinical dosages in use to abate this potential danger. With the dosage reduction, however, comes an inevitable sacrifice in image quality. In this paper, the latter problem is addressed by first modeling the noise that arises as a result of this dosage reduction. It is welLKnown that this noise is signal-dependent and PoissoNDistributed. A model for this type of noise in image sequences is formulated and the commonly utilized noise model for single images is shown to be obtainable from the new model. We propose stochastic temporal filtering techniques to enhance clinical fluorosocopy sequences corrupted by quantum mottle. The temporal versions of these niters as developed in this paper are more suitable for filtering image sequences, as correlations along the time axis can be utilized. For these dynamic sequences, the problem of displacement field estimation is treated in conjunction with the filtering stage to ensure that the temporal correlations are taken along the direction of motion to prevent object blur. {\textcopyright} 1993 IEEE},
author = {Chan, C.L. and Katsaggelos, A.K. and Sahakian, A.V.},
doi = {10.1109/42.241890},
issn = {02780062},
journal = {IEEE Transactions on Medical Imaging},
number = {3},
pages = {610--621},
publisher = {IEEE},
title = {{Image sequence filtering in quantum-limited noise with applications to low-dose fluoroscopy}},
url = {http://ieeexplore.ieee.org/document/241890/},
volume = {12},
year = {1993}
}

Clinical angiography—the procedure of acquiring radiographic (fluoroscopic) image sequences of patients from x-ray based medical systems—has unquestionably aided cardiologists in their assessment of coronary disease. During such trials, however, literally hundreds of x-ray images are gathered, thereby putting these patients and particularly the medical staff at risk. It is desirable to lower the clinical dosages in use to abate this potential danger. With the dosage reduction, however, comes an inevitable sacrifice in image quality. In this paper, the latter problem is addressed by first modeling the noise that arises as a result of this dosage reduction. It is welLKnown that this noise is signal-dependent and PoissoNDistributed. A model for this type of noise in image sequences is formulated and the commonly utilized noise model for single images is shown to be obtainable from the new model. We propose stochastic temporal filtering techniques to enhance clinical fluorosocopy sequences corrupted by quantum mottle. The temporal versions of these niters as developed in this paper are more suitable for filtering image sequences, as correlations along the time axis can be utilized. For these dynamic sequences, the problem of displacement field estimation is treated in conjunction with the filtering stage to ensure that the temporal correlations are taken along the direction of motion to prevent object blur. © 1993 IEEE

@article{Serafim1993,
abstract = {In this paper, an adaptive regularized recursive displacement estimation algorithm is presented. An estimate of the displacement vector field (DVF) is obtained by minimizing the linearized displaced frame difference (DFD) using v subsets (submasks) of a set of points that belong to a causal neighborhood (mask) around the working point. Assuming that the displacement vector is constant at all points inside the mask, v systems of equations are formed based on the corresponding submasks. A set theoretic regularization approach is followed for solving this system of equations by using information about the noise and the solution. An expression for the variance of the linearization error is derived in quantifying the information about the noise. Prior information about the solution is incorporated into the algorithm using a causal oriented smoothness constraint (OSC) which also provides a spatially adaptive prediction model for the estimated DVF. It is shown that certain existing regularized recursive algorithms are special cases of the proposed algorithm, if a single mask is considered. Based on experiments with typical videoconferencing scenes, the improved performance of the proposed algorithm with respect to accuracy, robustness to occlusion and smoothness of the estimated DVF is demonstrated. {\textcopyright} 1993 IEEE},
author = {Efstratiadis, S.N. and Katsaggelos, A.K.},
doi = {10.1109/83.236533},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
month = {jul},
number = {3},
pages = {341--352},
title = {{An adaptive regularized recursive displacement estimation algorithm}},
url = {http://ieeexplore.ieee.org/document/236533/},
volume = {2},
year = {1993}
}

In this paper, an adaptive regularized recursive displacement estimation algorithm is presented. An estimate of the displacement vector field (DVF) is obtained by minimizing the linearized displaced frame difference (DFD) using v subsets (submasks) of a set of points that belong to a causal neighborhood (mask) around the working point. Assuming that the displacement vector is constant at all points inside the mask, v systems of equations are formed based on the corresponding submasks. A set theoretic regularization approach is followed for solving this system of equations by using information about the noise and the solution. An expression for the variance of the linearization error is derived in quantifying the information about the noise. Prior information about the solution is incorporated into the algorithm using a causal oriented smoothness constraint (OSC) which also provides a spatially adaptive prediction model for the estimated DVF. It is shown that certain existing regularized recursive algorithms are special cases of the proposed algorithm, if a single mask is considered. Based on experiments with typical videoconferencing scenes, the improved performance of the proposed algorithm with respect to accuracy, robustness to occlusion and smoothness of the estimated DVF is demonstrated. © 1993 IEEE

@article{chan1993recursive,
abstract = {Quantum noise in image sequences arises in a wide array of applications including medical and astronomical images and remote sensing. It is an undesirable artifact caused by the unavailability or intentional depletion of X-ray and/or light photons necessary for imaging. In this paper, we develop a recursive sliding window, locally linear minimum mean squared error motion-compensated temporal filter for the enhancement of image sequences corrupted by this type of noise. The filter is derived from the formulation of a noise model which describes the underlying physical processes of quantum noise. The recursive implementation of this filter will provide an intuitive manner for addressing subsequent image frames as new observations when acquired under quantum-limited conditions. Experimental results are provided which show the effectiveness of the proposed estimator on both sequences simulated with quantum noise and on real clinical sequences containing natural quantum mottle. {\textcopyright} 1993 Academic Press. All rights reserved.},
author = {Chan, Cheuk L. and Katsaggelos, Aggelos K. and Sahakian, Alan V.},
doi = {10.1006/jvci.1993.1032},
issn = {10473203},
journal = {Journal of Visual Communication and Image Representation},
month = {dec},
number = {4},
pages = {349--363},
publisher = {Academic Press},
title = {{Recursive Locally Linear MMSE Motion-Compensated Image Sequence Filtering under Quantum-Limited Conditions}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1047320383710321},
volume = {4},
year = {1993}
}

Quantum noise in image sequences arises in a wide array of applications including medical and astronomical images and remote sensing. It is an undesirable artifact caused by the unavailability or intentional depletion of X-ray and/or light photons necessary for imaging. In this paper, we develop a recursive sliding window, locally linear minimum mean squared error motion-compensated temporal filter for the enhancement of image sequences corrupted by this type of noise. The filter is derived from the formulation of a noise model which describes the underlying physical processes of quantum noise. The recursive implementation of this filter will provide an intuitive manner for addressing subsequent image frames as new observations when acquired under quantum-limited conditions. Experimental results are provided which show the effectiveness of the proposed estimator on both sequences simulated with quantum noise and on real clinical sequences containing natural quantum mottle. © 1993 Academic Press. All rights reserved.

@article{yang1993regularized,
abstract = {The block discrete cosine transform (BDCT) is by far the most widely used transform for the compression of both still and sequences of images. High compression ratios are usually achieved by discarding information about the BDCT coefficients that is considered unimportant and yield images that exhibit the visually annoying blocking artifact. In this paper reconstruction of images from incomplete BDCT data is examined. The problem is formulated as one of regularized image recovery. According to this formulation, the image in the decoder is reconstructed by using not only the transmitted data but also prior knowledge about the smoothness of the original image, which complements the transmitted data. Two methods are proposed for solving this regularized recovery problem. The first is based on the theory of projections onto convex sets (POCS) while the second is based on the constrained least squares (CLS) approach. For the POCS-based method, a new constraint set is defined that conveys smoothness information not captured by the transmitted BDCT coefficients, and the projection onto it is computed. For the CLS method an objective function is proposed that captures the smoothness properties of the original image. Iterative algorithms are introduced for its minimization. Experimental results are presented that demonstrate that with the regularized reconstruction it is possible to drastically reduce the blocking artifact and improve the performance using both subjective and objective metrics of traditional decoders, which use the transmitted BDCT coefficients only. {\textcopyright} 1993 IEEE},
author = {{Yongyi Yang} and Galatsanos, Nikolas P. N.P. and Katsaggelos, Aggelos K. A.K. and Yang, Yongyi and Galatsanos, Nikolas P. N.P. and Katsaggelos, Aggelos K. A.K.},
chapter = {421},
doi = {10.1109/76.260198},
isbn = {10518215},
issn = {15582205},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
number = {6},
pages = {421--432},
publisher = {IEEE},
title = {{Regularized Reconstruction to Reduce Blocking Artifacts of Block Discrete Cosine Transform Compressed Images}},
url = {http://ieeexplore.ieee.org/document/260198/},
volume = {3},
year = {1993}
}

The block discrete cosine transform (BDCT) is by far the most widely used transform for the compression of both still and sequences of images. High compression ratios are usually achieved by discarding information about the BDCT coefficients that is considered unimportant and yield images that exhibit the visually annoying blocking artifact. In this paper reconstruction of images from incomplete BDCT data is examined. The problem is formulated as one of regularized image recovery. According to this formulation, the image in the decoder is reconstructed by using not only the transmitted data but also prior knowledge about the smoothness of the original image, which complements the transmitted data. Two methods are proposed for solving this regularized recovery problem. The first is based on the theory of projections onto convex sets (POCS) while the second is based on the constrained least squares (CLS) approach. For the POCS-based method, a new constraint set is defined that conveys smoothness information not captured by the transmitted BDCT coefficients, and the projection onto it is computed. For the CLS method an objective function is proposed that captures the smoothness properties of the original image. Iterative algorithms are introduced for its minimization. Experimental results are presented that demonstrate that with the regularized reconstruction it is possible to drastically reduce the blocking artifact and improve the performance using both subjective and objective metrics of traditional decoders, which use the transmitted BDCT coefficients only. © 1993 IEEE

@article{katsaggelos1993general,
abstract = {In this paper, we provide a general framework for performing linear shift-invariant within channel and shift varying across channels processing of stationary multi-channel (MC) signals. Emphasis is given on the restoration of degraded signals. We show that, by utilizing the special structure of semi-block circulant and block diagonal matrices, MC signal processing can be easily carried out in the frequency domain. The generalization of many frequency domain single-channel (SC) signal processing techniques to the MC case is presented. We show that in MC signal processing each frequency component of a signal and system is respectively represented by a small vector and a matrix (of size equal to the number of channels), while in SC signal processing each frequency component in both cases is a scalar. {\textcopyright} 1993 IEEE},
author = {Katsaggelos, A.K. and Lay, K.T. and Galatsanos, N.P.},
doi = {10.1109/83.236528},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
month = {jul},
number = {3},
pages = {417--420},
publisher = {IEEE},
title = {{A general framework for frequency domain multi-channel signal processing}},
url = {http://ieeexplore.ieee.org/document/236528/},
volume = {2},
year = {1993}
}

In this paper, we provide a general framework for performing linear shift-invariant within channel and shift varying across channels processing of stationary multi-channel (MC) signals. Emphasis is given on the restoration of degraded signals. We show that, by utilizing the special structure of semi-block circulant and block diagonal matrices, MC signal processing can be easily carried out in the frequency domain. The generalization of many frequency domain single-channel (SC) signal processing techniques to the MC case is presented. We show that in MC signal processing each frequency component of a signal and system is respectively represented by a small vector and a matrix (of size equal to the number of channels), while in SC signal processing each frequency component in both cases is a scalar. © 1993 IEEE

@article{Aggelos1992,
abstract = {In this paper mesh, pyramid, and mesh of pyramids implementations of an iterative image restoration algorithm are proposed. These implementations are based on a single step regularized iterative restoration algorithm. Area-time bounds on the proposed implementations are established. The efficiency of the proposed VLSI algorithms is evaluated by comparing the established bounds against lower bounds on AT 2 , where A is the area of the VLSI chip and T is its computation time.},
author = {KATSAGGELOS, A.K. and KUMAR, S.P.R. and SARRAFZADEH, M.},
doi = {10.1142/S0218126692000179},
issn = {0218-1266},
journal = {Journal of Circuits, Systems and Computers},
month = {sep},
number = {03},
pages = {265--280},
title = {{VLSI ARCHITECTURES FOR ITERATIVE IMAGE RESTORATION}},
url = {https://www.worldscientific.com/doi/abs/10.1142/S0218126692000179},
volume = {02},
year = {1992}
}

In this paper mesh, pyramid, and mesh of pyramids implementations of an iterative image restoration algorithm are proposed. These implementations are based on a single step regularized iterative restoration algorithm. Area-time bounds on the proposed implementations are established. The efficiency of the proposed VLSI algorithms is evaluated by comparing the established bounds against lower bounds on AT 2 , where A is the area of the VLSI chip and T is its computation time.

@article{Kang1992,
author = {Kang, M.G. and Katsaggelos, A.K.},
doi = {10.1109/78.157234},
issn = {1053587X},
journal = {IEEE Transactions on Signal Processing},
number = {9},
pages = {2329--2334},
title = {{Simultaneous iterative image restoration and evaluation of the regularization parameter}},
url = {http://ieeexplore.ieee.org/document/157234/},
volume = {40},
year = {1992}
}

@article{galatsanos1992methods,
abstract = {The application of regularization to ill-conditioned problems necessitates the choice of a regularization parameter which trades fidelity to the data with smoothness of the solution. The value of the regularization parameter depends on the variance of the noise in the data. In this paper the problem of choosing the regularization parameter and estimating the noise variance in image restoration is examined. An error analysis based on an objective mean square error (MSE) criterion is used to motivate regularization. Two new approaches for choosing the regularization parameter and estimating the noise variance are proposed. The proposed and existing methods are compared and their relation to linear minimum mean square error (LMMSE) filtering is examined. Experiments are presented that verify the theoretical results. {\textcopyright} 1992 IEEE},
author = {Galatsanos, N.P. and Katsaggelos, A.K.},
doi = {10.1109/83.148606},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
month = {jul},
number = {3},
pages = {322--336},
title = {{Methods for choosing the regularization parameter and estimating the noise variance in image restoration and their relation}},
url = {http://ieeexplore.ieee.org/document/148606/},
volume = {1},
year = {1992}
}

The application of regularization to ill-conditioned problems necessitates the choice of a regularization parameter which trades fidelity to the data with smoothness of the solution. The value of the regularization parameter depends on the variance of the noise in the data. In this paper the problem of choosing the regularization parameter and estimating the noise variance in image restoration is examined. An error analysis based on an objective mean square error (MSE) criterion is used to motivate regularization. Two new approaches for choosing the regularization parameter and estimating the noise variance are proposed. The proposed and existing methods are compared and their relation to linear minimum mean square error (LMMSE) filtering is examined. Experiments are presented that verify the theoretical results. © 1992 IEEE

@article{Joon1992,
abstract = {An edge detection algorithm using multi-state adaptive linear neurons (ADALINES) is presented. Although the tri-state ADALINE is only considered in this work, general multi-state input vectors with extreme values are shown to be linearly separable from the rest of the vectors with the same dimension. The input state of each ADALINE is defined using the local mean in a predefined mask. In addition to the binary input states ± 1, the 0 input state is introduced for controlling the noise effect. If the input pattern matches one of the predefined edge patterns, the corresponding pixel is detected as an edge pixel. Experimental results are shown where the proposed detector is compared with both the Canny and LOG edge detectors. {\textcopyright} 1992.},
author = {Paik, Joon K. and Brailean, James C. and Katsaggelos, Aggelos K.},
doi = {10.1016/0031-3203(92)90122-Y},
issn = {00313203},
journal = {Pattern Recognition},
keywords = {Edge detection,Linear neural networks,Pattern recognition},
month = {dec},
number = {12},
pages = {1495--1504},
title = {{An edge detection algorithm using multi-state adalines}},
url = {https://linkinghub.elsevier.com/retrieve/pii/003132039290122Y},
volume = {25},
year = {1992}
}

An edge detection algorithm using multi-state adaptive linear neurons (ADALINES) is presented. Although the tri-state ADALINE is only considered in this work, general multi-state input vectors with extreme values are shown to be linearly separable from the rest of the vectors with the same dimension. The input state of each ADALINE is defined using the local mean in a predefined mask. In addition to the binary input states ± 1, the 0 input state is introduced for controlling the noise effect. If the input pattern matches one of the predefined edge patterns, the corresponding pixel is detected as an edge pixel. Experimental results are shown where the proposed detector is compared with both the Canny and LOG edge detectors. © 1992.

@article{Serafim1992,
abstract = {In this paper, an approach for the constrained recursive estimation of the displacement vector field (DVF) in image sequences is presented. An estimate of the displacement vector at the working point is obtained by minimizing the linearized displaced frame difference based on a set of observations that belong to a causal neighborhood (mask). An expression for the variance of the linearization error (noise) is obtained. Because the estimation of the DVF is an ill-posed problem, the solution is constrained by considering an autoregressive (AR) model for the DVF. This AR model is first considered stationary, according to which the two components of the DVF are uncorrelated and each component is modeled by a 2-D discrete Markov sequence. A nonstationary AR model of the DVF is also considered by spatially adapting the model coefficients using a weighted LMS algorithm. Additional information about the solution is incorporated into the algorithm using a causal “oriented smoothness” constraint. Based on the above formulation, a set theoretic regularization approach is followed that results in a weighted constrained least-squares estimation of the DVF. The proposed algorithm shows an improved performance with respect to accuracy, robustness to occlusion, and smoothness of the estimated DVF when applied to typical videoconferencing scenes. {\textcopyright} 1992 IEEE},
author = {Efstratiadis, Serafim N. and Efstratiadis, Serafim N. and Katsaggelos, Aggelos K.},
doi = {10.1109/76.168901},
issn = {15582205},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
number = {4},
pages = {334--346},
title = {{Nonstationary AR modeling and constrained recursive estimation of the displacement field}},
volume = {2},
year = {1992}
}

In this paper, an approach for the constrained recursive estimation of the displacement vector field (DVF) in image sequences is presented. An estimate of the displacement vector at the working point is obtained by minimizing the linearized displaced frame difference based on a set of observations that belong to a causal neighborhood (mask). An expression for the variance of the linearization error (noise) is obtained. Because the estimation of the DVF is an ill-posed problem, the solution is constrained by considering an autoregressive (AR) model for the DVF. This AR model is first considered stationary, according to which the two components of the DVF are uncorrelated and each component is modeled by a 2-D discrete Markov sequence. A nonstationary AR model of the DVF is also considered by spatially adapting the model coefficients using a weighted LMS algorithm. Additional information about the solution is incorporated into the algorithm using a causal “oriented smoothness” constraint. Based on the above formulation, a set theoretic regularization approach is followed that results in a weighted constrained least-squares estimation of the DVF. The proposed algorithm shows an improved performance with respect to accuracy, robustness to occlusion, and smoothness of the estimated DVF when applied to typical videoconferencing scenes. © 1992 IEEE

@article{Aggelos1992a,
author = {Aggelos, K Katsaggelos and Richard, J Mammone},
journal = {Journal of Visual Communication and Image Representation},
pages = {305--306 , publisher = Academic Press},
title = {{Image restoration and reconstruction Guest Editors' comments}},
volume = {3},
year = {1992}
}

@article{paik1992image,
abstract = {In this paper a modified Hopfield neural network model for regularized image restoration is presented. The proposed network allows negative autoconnections for each neuron. A set of algorithms using the proposed neural network model is presented, with various updating modes: i) sequential updates, ii) n-simultaneous updates, and iii) partially asynchronous updates. The sequential algorithm is shown to converge to a local minimum of the energy function after a finite number of iterations. This local minimum is defined in terms of a unit distance neighborhood in the discrete intensity value space. Since an algorithm which updates all n neurons simultaneously is not guaranteed to converge, a modified algorithm is presented, which is called a greedy algorithm. Although the greedy algorithm is not guaranteed to converge to a local minimum, the 1<inf>1</inf> norm of the residual at a fixed point is bounded. It is also shown that the upper bound on the 1<inf>1</inf> norm of the residual can be made arbitrarily small by using an appropriate step size. Finally, a partially asynchronous algorithm is presented, which allows a neuron to have a bounded time delay to communicate with other neurons. Such an algorithm can eliminate the synchronization overhead of synchronous algorithms. It is shown that the /<inf>1</inf> norm of the residual at the fixed point of this algorithm increases as the upper bound on the delay increases. Experimental results are shown testing and comparing the proposed algorithms. {\textcopyright} 1992 IEEE},
author = {Paik, J.K. and Katsaggelos, A.K.},
doi = {10.1109/83.128030},
issn = {10577149},
journal = {IEEE Transactions on Image Processing},
number = {1},
pages = {49--63},
publisher = {IEEE},
title = {{Image restoration using a modified Hopfield network}},
url = {http://ieeexplore.ieee.org/document/128030/},
volume = {1},
year = {1992}
}

In this paper a modified Hopfield neural network model for regularized image restoration is presented. The proposed network allows negative autoconnections for each neuron. A set of algorithms using the proposed neural network model is presented, with various updating modes: i) sequential updates, ii) n-simultaneous updates, and iii) partially asynchronous updates. The sequential algorithm is shown to converge to a local minimum of the energy function after a finite number of iterations. This local minimum is defined in terms of a unit distance neighborhood in the discrete intensity value space. Since an algorithm which updates all n neurons simultaneously is not guaranteed to converge, a modified algorithm is presented, which is called a greedy algorithm. Although the greedy algorithm is not guaranteed to converge to a local minimum, the 11 norm of the residual at a fixed point is bounded. It is also shown that the upper bound on the 11 norm of the residual can be made arbitrarily small by using an appropriate step size. Finally, a partially asynchronous algorithm is presented, which allows a neuron to have a bounded time delay to communicate with other neurons. Such an algorithm can eliminate the synchronization overhead of synchronous algorithms. It is shown that the /1 norm of the residual at the fixed point of this algorithm increases as the upper bound on the delay increases. Experimental results are shown testing and comparing the proposed algorithms. © 1992 IEEE