Efficient stereo matching based on a new confidence metric

In this paper, we propose a new confidence metric for efficient stereo matching. To measure the confidence of a stereo match, we refer to the curvatures around the two minimum costs of a cost curve, the size of aggregation kernel, and the occlusion information. Using the proposed confidence metric, we then design a weighted median filter, in order to refine the initially estimated disparities with a small aggregation kernel. In the design of weighted median filter, we overcome the performance degradation due to a small kernel size by utilizing the filter information of previously processed pixels. It is found that the performance of the proposed stereo matching algorithm is competitive to the other existing local algorithms even with a small size of aggregation kernel.