Dynamic search range using sparse disparity map for fast stereo matching

In this paper, we have suggested a dynamic search range method to reduce a matching cost of the local stereo matching process basing on the sparse representation theory. The suggested method analyzes sparse disparity map, generating the valid search range information and then using the analyzed information skip the region where the matching blocks are not likely to be found in a dense disparity map acquisition step. Also considering for the solution of accuracy decline, we have offered some parameters to adjust speed and accuracy trade-off. Through experimental result, we have showed that the suggested algorithm reduces computation time 40% to 70% with a acceptable accuracy loss. The proposed algorithm can be applied to any local stereo matching algorithm. Also our algorithm is designed to be hardware friendly by lowering the hardware resource requirement and memory bandwidth.