A Window-Based Adaptive Correspondence Search Algorithm Using Mean Shift and Disparity Estimation

Aiming at the problem of low efficiency and unsatisfactory matching of uniform texture regions in binocular stereo vision, we propose a rapid window-based adaptive correspondence search algorithm using mean shift and disparity estimation. Color aggregation is firstly carried out to the reference image and the target image through mean shift method in order to obtain images with low dynamic color range. Then we make disparity estimation to the pre-processed two images and compute disparities of uniform texture regions. Finally, adaptive window matching is completed and exact depth map is achieved through similarity computation and window-based support aggregation. Experimental results show that our algorithm is more efficient and keeps smooth disparity better than the prior window method.