Dense Disparity Computing Method Based on Mesh Aggregation and Snake Optimization for Stereo Vision

Abstract

In order to generate continuous and dense disparity images, a stereo matching method based on mesh aggregation and Snake optimization is proposed in this article. First, the reference pixels are obtained, so as to improve the suppression effect of the brightness difference in Census transform and improve the accuracy of initial matching cost calculation. Second, the image is divided by SLIC super pixel segmentation method, and the neighborhood pixels are searched according to the mesh search in the region, and the matching cost of these pixels are aggregated together according to the corresponding weight to complete cost aggregation of the pixels to be matched. Third, the Snake algorithm is used in optimizing the boundary of the disparity region. Eight classes of images on the Middlebury platform are selected as the test images, and the four algorithms on the Middlebury platform are selected as reference algorithms to carry out the experimental research. The experimental results show that proportion to bad pixels is low and disparity is continuous and dense on the disparity image calculated by the algorithm proposed in this article. Performance of the proposed method is close to LocalExp algorithm which is the best on the Middlebury platform, and the proposed method can be better applied in the stereo vision.