A Fast Cooperative Algorithm for Stereo Matching

This paper presents a fast stereo algorithm for obtaining disparity maps efficiently. We use a 3D model for storing and computing the depth map. The initial matching by intensity similarity is very fast by using the computational optimization. At the improving matching reliability step, two assumptions that were originally proposed by Marr and Poggio are adopted: uniqueness and continuity. It means there is only one unique depth value for each pixel in the disparity maps and the depths of most pixels are continuous. We presents a novel restrict function using inhibition area without iteration and enhanced by the initial matching result for keeping the sharpness. We present the results comparing with the well-known algorithm based on SMP which confirm that our algorithm is better than SMP on accuracy.