Motion Segmentation by Robust Fitting and Model Selection
Kenichi Kanatani
Gunma University

Time and Place

3305 Newell-Simon Hall
Refreshments 3:15 pm
Talk 3:30 pm

Abstract

We dramatically improve the performance of the Costeira-Kanade algorithm for motion segmentation from feature tacking data. The key idea is to view the Costeira-Kanade algorithm as a pure mathematical theorem for separating subspaces in a high-dimensional space of motion history, which is independent of the Tomasi-Kanade factorization for 3-D reconstruction. We present a robust segmentation algorithm by introducing such techniques as statistical optimization, dimension correction, model selection using the geometric AIC, and least-median fitting. We infer the number of independently moving objects by estimating the matrix rank using the geometric AIC and the geometric MDL. A prominent feature of our algorithm is that no parameters that need to be empirically adjusted are involved. We evaluate the reliability of the resulting segmentation using numerical simulations and real image examples.

Speaker Biography

Kenichi Kanatani received his B.S., M.S, and Ph.D. in applied mathematics from the University of Tokyo, Japan, in 1972, 1974, and 1979, respectively.

He joined the Department of Computer Science, Gunma University, Kiryu, Japan, in April 1979 as Assistant Professor. He became Associate Professor and Professor there in April 1983 and April 1988, respectively.

He was a visiting researcher at the University of Maryland, U.S.A., the University of Copenhagen, Denmark, the University of Oxford, U.K., and INIRA at Rhone Alpes, France.

He is the author of ``Group-Theoretical Methods in Image Understanding'' (Springer, 1990), ``Geometric Computation for Machine Vision'' (Oxford University Press, 1993) and ``Statistical Optimization for Geometric Computation: Theory and Practice'' (Elsevier Science, 1996).

His his research career started with studies of theoretical continuum mechanics (elasticity, plasticity, and fluid) and its application to mechanics of granular materials such as powder and soil, but his research interested has shifted to mathematical analysis of images and 3-D reconstruction from images. Currently, he is devoted to mathematical analysis of statistical reliability of computer vision and optimization procedures. He is a Senior Member of IEEE.

Speaker Appointments

For appointments, please contact the host, Takeo Kanade (tk@cs.cmu.edu).