Dense Stereo Matching Algorithm Based On Image Segmentation

A dense stereo matching algorithm is proposed based on image segmentation. This algorithm combines the gray-gradient algorithm and the zero-mean normalized cross-correlation (ZNCC) algorithm to generate matching cost. The SLIC (Simple Liner Iterative Cluster) algorithm is used for image segmentation. A method based disparity map and superpixels is proposed to update the matching cost. At the disparity post-processing stage, the LRC (Left Right Check), hole filling and cross adaptive window weighted median filtering methods arc used to reduce the error matching rate of the disparity map. The performance evaluation experiments on four Middlebury stereo pairs demonstrate that the proposed algorithm achieves an average error matching rate of 4.99% Key words machine vision; stereo matching algorithm; matching cost computation method fusion; cross adaptive window weighted median filtering