Sparsity, Statistical and Physical Models for Computer Vision
  • FENS
  • Sparsity, Statistical and Physical Models for Computer Vision

You are here


Sparsity, Statistical and Physical Models for Computer Vision, Medical Image Analysis and Graphics

Dimitris Metaxas
Distinguished Professor of Computer Science
Director of Computational Biomedicine, Imaging, and Modeling (CBIM) Center
Rutgers, The State University of New Jersey

Friday, May 25, 14:40-15:30, FENS G029

During this talk we will present a computational framework that was developed at the Center for Computational Biomedicine, Imaging and Modeling (CBIM) at Rutgers University  and is based on novel developments in statistical and physical models. Our approach has applications to computer vision, medical image analysis and computer graphics.

In the first part of the talk we will present recent advances in deformable models have lead to new classes of methods that borrow the best features form level sets as well as traditional parametric deformable models. This new class of such models we term, Metamorphs,  integrates shape, intensity  and texture. These new models can be used in medical segmentation and registration where organ boundaries are fuzzy and with no assumptions on the noise distribution. Applications include cancer and cardiac detection and reconstruction that has lead to a cardiac analysis system from MRI-tagged and CT data. Using, these methods for cardiac modeling, I will present some recent advances in fluid modeling which allow the simulation of coupled fluid-deformable object interaction at very fine scales. These new fluid models have been used for the cardiac blood flow analysis from CT data

In the second part of the talk we will present a new learning formulation called structured sparsity, which is a natural extension of the standard sparsity concept in statistical learning and compressive sensing. By allowing arbitrary structures on the feature set, this concept generalizes to the group sparsity idea. A general theory is developed for learning with structured sparsity, based on the notion of coding complexity associated with the structure. We will show that it can be applied to different applications such as sparse learning, compressive sensing and medical imaging.

Finally, we will present the use of stochastic deformable models to problems related to nonverbal communication analysis and in particular their use in real time facial tracking , expressions analysis for deception and other applications.

Dr. Dimitris Metaxas is a Professor II (Distinguished) in the Division of Computer and Information Sciences. He is directing the Center for Computational Biomedicine, Imaging and Modeling (CBIM). From September 1992 to September 2001 he was a Professor in the Computer and Information Science Department of the University of Pennsylvania and Director of the VAST Lab. Prof. Metaxas received a Diploma in Electrical Engineering from the National Technical University of Athens Greece in 1986, an M.Sc. in Computer Science from the University of Maryland, College Park in 1988, and a Ph.D.  in Computer Science from the University of Toronto, Ontario, Canada in 1992.

Dr. Metaxas has been conducting research towards the development of formal methods upon which both computer vision, computer graphics and medical imaging can advance synergistically.  In computer vision, he works on the simultaneous segmentation and fitting of complex objects, shape representation, deterministic and statistical object tracking, sparse methods for segmentation and restoration, learning and ASL, gesture recognition and human activity analysis.  In the area of biomedical applications new methods have been developed for material modeling and shape estimation of internal body parts (e.g., heart, brain and lungs) from MRI, SPAMM and CT data, a framework for linking the anatomical and physiological models of the human body and deformable models suitable for segmentation, registration and reconstruction. Recently, new methods based on sparsity have been developed for segmentation and image reconstruction.  In computer graphics, he introduced the Navier-Stokes methodology for Fluids, based on which the water scenes in the movie Antz were created in 1998. For this work, his student Nick Foster won an Academy Award in 1998. Since then, he is working on new techniques for modeling fluid phenomena, and control theoretic techniques for automating and improving the animation of articulated (e.g., humans) objects. Dr. Metaxas has published over 350 research articles in these areas and has graduated 33 PhD students. The above research has been funded by NSF, NIH, ONR, AFOSR, DARPA, HSARPA and the ARO.

Dr. Metaxas has published a book on his research activities titled ``Physics-based deformable models: Applications to computer vision, graphics and medical imaging'' which was published by Kluwer Academic. He organized the first IEEE Workshop on Physics-Based Modeling in Computer Vision, he is on the Editorial Board of Medical Imaging, as Associate Editor of GMOD, and CAD and is a Co-Editor of the Special Issue of Computer Vision and Image Understanding on Physics-Based Modeling and Reasoning. Dr. Metaxas has received several  paper awards for research and he has several patents. He was awarded a Fulbright Fellowship in 1986, is a recipient of an NSF Research Initiation and Career awards, an ONR YIP, and is a Fellow of the American Institute of Medical and Biological Engineers, and a member of ACM and IEEE. He has been involved with the organization of several top conferences in vision (Program Chair of ICCV 2007 and General Chair of ICCV 2011), medical imaging (General Chair of MICCAI 2008) and graphics (Senior Program Chair for SCA 2007).