Enhancing the volumetric approach to stereo matching

Describes a volumetric stereo algorithm for obtaining disparity maps and identifying the presence of occlusions. It works in the (row × column × disparity range) space, initializing the voxels with a measure of the similarity between the stereo pair they represent (similarity phase) and refining these values through an iterative process which inhibits all voxels but one, placed along the same line of sight (competition phase). By ceasing the updating of those voxels that showed little promise, it was possible to enhance the time performance without influencing the results in all tested examples. Also, to obtain smoother maps, a dynamic programming process is applied to obtain the final solution