Bayesian Analysis of Image Sequences: Detection and Tracking of Motion Boundaries

David Fleet
Xerox PARC and Queen's University

Abstract

Motion analysis concerns the estimation and recognition of motion from image sequences. It is useful for scene segmentation, estimating 3D surface structure and camera motion, and for the detection and tracking of objects, such as people. A long-standing problem in motion analysis has been the detection and estimation of motion in the neighborhoods of surface boundaries, where motion in the image is discontinuous and occlusions cause image structure to appear or disappear from one image to the next. Although these "motion boundaries" are often viewed as a source of noise for current motion estimation techniques, we can also view them as a rich source of information about the location of surface boundaries and the depth ordering of surfaces at these locations.

We propose a Bayesian framework for representing and estimating image motion in terms of multiple motion models, including both smooth motion and local motion discontinuity models. We compute the posterior probability distribution over models and model parameters, given the image data, using discrete samples and a particle filter for propagating beliefs through time. In this talk I will introduce the problem then describe our Bayesian framework, including the generative models, the likelihood computation, the particle filter, and a mixture model prior from which samples are drawn. I will also present some recent experimental results.

About the Speaker

David Fleet is a research scientist at the Xerox Palo Alto Research Center (PARC), and an Associate Professor in the Department of Computer Science at Queen's University, Kingston, Canada. After receiving a PhD in Computer Science from the University of Toronto in 1991, Dr. Fleet joined the Department of Computer Science at Queen's University, with cross-appointments to the Departments of Psychology, and Electrical Engineering. In 1999 he joined the Digital Video Analysis Group at Xerox PARC, where his research is focused on motion analysis in computer vision. Dr. Fleet was awarded an Alfred P. Sloan Research Fellowship for his research on biological vision. In 1999 his paper with Michael Black on probabilistic detection and tracking of motion discontinuities received Honorable Mention for the Marr Prize at the International Conference on Computer Vision. His research interests include computer vision, image processing, visual perception, and visual neuroscience. He has published research articles and one book on various topics including the estimation of optical flow and stereoscopic disparity, probabilistic methods in motion analysis, modeling appearance changes in image sequences, non-Fourier motion and stereo perception, and the neural basis of stereo vision.
bac-coordinators@cs.stanford.edu
Back to the Colloquium Page
Last modified: Tue Sep 21 18:59:46 PDT 1999