Abstract
This paper introduces a very simple but high-quality stereo algorithm. Generally in stereo algorithms, window matching is used for stable results. A larger window introduces more-stable results. But a large window often contains objects with different disparities. False matching often occurs in these areas because of illegal high evaluation values that result from occlusion. This algorithm avoids false matching by setting an upper limit for evaluation values to restrict illegal high evaluation values. Experimental results show improved matching accuracy.