Semi-Global Stereo Matching Algorithm Based on Minimum Spanning Tree

The classical semi-global matching algorithm can't use of the pixels around the points to be matched efficiently in the matching cost aggregation process, which will cause mis-match in the weak texture and occlusion region of the image. To solve this problem, the minimum spanning tree is added to the matching cost aggregation process of semi-global stereo matching algorithm, and the matching cost is calculated by dynamic programming along four paths. Solving the matching cost after aggregation within the maximum allowable range of parallax, Finding the parallax that minimizes the matching cost as the apparent difference of the point to be matched. The experimental results show that the Minimum Spanning Tree can be used to treat the pixels around the matching points efficiently and fully, it overcomes the problem that the semi-global matching algorithm can't make full use of the pixels of the image, compared with the original algorithm, the matching accuracy is higher and the aggregation path of matching cost is reduced from 16 to 4, which improves the matching efficiency.