May 10th, 2002

Title : Stereoscopic Segmentation

Hailin Jin
Department of Electrical Engineering
Washington University

We cast the problem of shape reconstruction of a scene as the global region segmentation of a collection of calibrated images. We assume that the scene is composed of a number of smooth surfaces and a background, both of which support smooth radiance functions. We formulate the problem in a variational framework, where the solution (both the shape and radiances) is a minimizer of a global cost functional which combines a geometric prior on shape, a smoothness prior on radiance and the data fitness. We estimate the shape and radiances via an alternating minimization: The radiances are computed as the solutions of partial differential equations defined on the surfaces and the background. The shape is estimated using a gradient descend flow, which is implemented using the level set methods. Our algorithm works for Lambertian scenes with smooth or constant radiances as well as fine homogeneous textures, which are known challenges to traditional stereo algorithms based on local correspondence.