Collective Stereopsis on the Hypercube

A cooperative algorithm for extracting disparity information from stereo image pairs has been implemented on the NCUBE hypercube computer. Software is written in C-langauge, using communication routines of the “Crystalline Operating System” CrOSIII designed at Caltech within the Caltech Concurrent Computation Program. Some tests have been done using Julesz’s random-dot stereograms. Although the software is reasonably versatile and can be easily adapted for different flavours of stereo algorithms, the method used is that presented by Marr and Poggio [REF.l]. As a preliminary stage for the stereo matching problem, a “filtering” program to extract physically meaningful primitives from images of a given scene has been written and tested on various types of images.