Reaction-diffusion Algorithm For Stereo Disparity Detection

摘要

The present paper proposes a novel stereo algorithm utilizing multi-sets of reaction-diffusion equations. The problem of detecting a stereo disparity map becomes the segmentation problem, in which the uniqueness assumption and the continuity assumption on disparity distribution are taken into account. A set of reaction-diffusion equations realizes the continuity assumption, while a mutual-inhibition mechanism among the multi-sets realizes the uniqueness one. In addition, each set of equations has a self-inhibition mechanism, which is necessary for the reaction-diffusion equations applied to stereo disparity detection. Performance of the proposed algorithm is evaluated for well-known test stereo images.