A fast parallel algorithm for stereovision

A new stereo algorithm based on local spatial filter operations and a fast coherence detection scheme is presented. The algorithm achieves performance similar to classical area-based approaches, but without the complicated hierarchical search structure typical for these approaches. The algorithm is fully parallel; disparity values and a corresponding verification count are calculated independently for each pixel. In addition, the network structure is able to fuse the left and right input images of the scene within its fusion range into a single cyclopean view.