Stochastic Approachres to Early Vision: Stereo Disparity
Sandip Chatterjee, Chienchung Chang · 2005
rl Ihe various task modules in early vision bear resemblances in processing visual cues. A successful model in one module can be adapted to other modules. In this paper, we investigate the applicability of stochastic models for computing disparity from a pair of images (stereo). Several computational theories for obtaining the maximum a posteriori estimate of stereo disparity map are discussed and experimental results are provided to demonstrate their effectivenesses. In addition, we discuss the problems caused by depth discontinuities as well as occlusion areas.