Research on Stereo Matching Algorithm Based on Improved Steady-State Matching Probability
Jianye Zhang, Yan Piao · Journal of Physics Conference Series · 2018
Aiming at the hole phenomenon generated by steady-state matching probability (SSMP) stereo matching algorithm in the test graph where the disparity is large,and the problem of correct disparity loss in the left disparity map caused by the error disparity in the right disparity map after using that algorithm. A stereo matching algorithm based on SSMP and semi-global matching (SGM) is proposed. Using the consistency detection method of the disparity maps obtained by SSMP and SGM are presented to fill the disparity map and the filling criterion based on hill-climbing algorithm is introduced. The experimental results demonstrate that the average bad pixels rates of the improved SSMP algorithm is reduced from 5.38% to 5.23% on the Middlebury benchmark. And the proposed algorithm can deal with the above problems well, improve the matching accuracy effectively and robustness.