Document Type

Technical Report

Publication Date






Technical Report Number



The goal of stereo vision is the recovery of depth information from the relative lateral displacements in the positions of objects within a pair of images taken from slightly differing viewpoints. While recent stereo matching techniques have yielded improvements in reliability and speed, most of these algorithms fall short of the real-time stereo matching requirements for navigation systems, robot vision, machine inspection, and other areas computer vision where rapid response is critical. Traditionally, matching algorithms have achieved high speeds through algorithm simplification and/or relied on custom hardware. The objective of our work has been the develop of a robust high-speed stereo matcher by exploiting parallel algorithms executing on general purpose SIMD machines. Our approach is based on several existing techniques dealing with the classification and evaluation of matches, the application of ordering constraints, and relaxation-based matching. The techniques have bene integrated and reformulated in terms of parallel execution on a theoretical SIMD machine. An ideal machine topology for executing this parallel algorithm is identified through complexity analysis. Feasibility is demonstrated by implementation on a commercially available SIMD machine, and its performance compared with that of the idealized machine. Sample results are shown for real and synthetic stereo pairs.


Permanent URL: