General and fast algorithm for disparity error detection and correction

General error detection and correction algorithm for disparity maps is developed to provide precise stereo matching results. A novel global order matching constraint was applied to post process the disparity map (but not in the disparity computation process) to detect cross regions in the disparity map. A zero cross correction algorithm was proposed to correct the erroneously matched positions. The algorithm was tested with a set of real stereo images and their corresponding disparity maps. The quality of the corrected disparity map was greatly improved. The method can be appended to various stereo matching algorithms as a general technique for post processing of disparity maps.