Iterative relaxational stereo matching based on adaptive support between disparities

摘要

An adaptive relaxation method for stereo matching using matching probability and compatibility coefficients between disparities is presented. Relaxation labeling is an iterative procedure which reduces local ambiguity and noise, and selects the best label among several possible choices using compatibility coefficient. Stereo matching can be considered as labeling problem which assigns unique matches to feature points in the image. The compatibility coefficient between disparity classes is determined from highly reliable matches in the initial matching result and quantitatively expressed. The matching results of neighbor pixels support the center pixel through compatibility coefficients and update its matching probability. As the true matching results are propagated, so false matches are corrected and have high probability values. The conventional matching algorithms permit a uniform support from different disparity regions and it causes the degradation phenomena on the discontinuities of object. The results of experiments show that the proposed adaptive supporting method reduces the degradations and obtains fast convergence.