ECE Seminar: On Structure and Motion
Monday, April 29, 2013 - 11:45am to 1:00pm
Guy Rosman, Department of Computer Science, Technion
One of the most important aspects of solving a problem in image, signal, or surface processing, is that of choosing an appropriate parametrization. This trivial observation can be seen in many forms, including both global parametrizations (such as the Hough and Fourier transforms) as well as local ones (such as Sparsity-based patch models and over-parameterized optical flow approaches). My research explores important cases in motion analysis and 3D reconstruction where the right selection of a local parametrization leads to simple and yet generic formulations that can be efficiently implemented. The first part of my talk relates to 3D motion understanding, where I demonstrate how articulated motion can be formulated as edge-preserving smoothing of Lie-group-valued images. The resulting generic algorithm obtains results comparable to those of domain specific tools, on 3D range data, at real-time speeds, and applies also to other inverse problems such as diffusion tensor imaging reconstruction, and direction diffusion. In the second part I show how structured light can be formulated as probability maximization with respect to the range image. This allows us to incorporate sparse priors for the surface into the non-linear reconstruction process itself, obtaining 3D reconstruction that is robust to low sensor exposure and motion artifacts. BIO: Guy Rosman is a PhD student in the department of Computer Science at the Technion.