Seminars/Colloquia/Invited Talks

Seminars

Ameesh Makadia

Structure from Motion without Correspondences

PLACE:	Clark 314
EVENT:	CIS Seminar Series
DATE:	June 6, 2006
TIME:	1:00 - 2:00

Abstract

A multitude of computer vision applications are linked by the fundamental problem of matching. Some examples are signal alignment for image mosaicing, point matching for laser scan alignment, silhouette matching for 3D shape retrieval, and feature matching for 3D motion estimation. This talk will touch on the application of harmonic analysis techniques for all of these examples, but the focus will be on the last one. We will present a novel approach for the correspondenceless estimation of 3D-motion directly from two images using the Radon transform, where the feasibility of any camera motion is computed by integrating over all feature pairs that satisfy the epipolar constraint. The main novelty is in the realization that this Radon integral can be expressed as a correlation of spherical functions. We propose a new algorithm to compute this integral using the spherical Fourier transform, which reduces the complexity of the Radon computation by a factor equal to the number of feature pairs processed. The strength of the algorithm is in avoiding a commitment to correspondences, thus being robust to erroneous feature detection, outliers, and multiple motions.

Brief biography

Ameesh Makadia is a PhD Candidate in the GRASP Laboratory in the Department of Computer and Information Science at the University of Pennsylvania. He is working with Dr. Kostas Daniilidis. His interests are structure-from-motion, motion estimation, and registration. His research focuses on correspondenceless vision techniques and the application of harmonic analysis theory to various vision problems.