A two-step stereo correspondence algorithm based on combination of feature-matching and region-matching

Stereo matching is a hot issue in the study of stereo vision. Due to the traditional region-matching algorithm has high error rate and large calculation amount, combined with the respective advantages of feature-based algorithm and region-based algorithm, a new two-step stereo correspondence algorithm based on combination of feature-matching and region-matching is proposed. First, use Harris corner detection algorithm to extract the corner, then higher accuracy correspondence points are obtained by feature matching, which are used to estimate the disparity search space. Second, in the region-matching part, use the calculated search space to get the overall dense disparity map. As for the unbelievable disparity forecast point caused by the blocked area, calculate its error energy and set appropriate threshold value to filter the unbelievable disparity point. The simulation results show that the algorithm has significantly improved matching speed and accuracy.