An architecture for stereo image matching

This paper addresses a modified dynamic programming parallel algorithm, and its associated architecture of temporal and spatial complexity of O(N), with N number of processing elements involved in computation. The modification consists of the algorithm adaptation to image interest point matching and architecture optimization. The parallel algorithm simulates the designed architecture, validates it functionally, and allows temporal performance prediction. The proposed algorithm of low spatial and temporal complexities is very suitable for real time on board applications (autonomous robot guidance, trajectory matching, blind vision guidance...).