An improved stereo matching algorithm based on corner points

The corners of Harris extracted based on the image gray gradient always have detection bias, because of the unclear gray gradient near the corners. In this paper, the `virtual corner´ is proposed, and the environmental difference also be introduced as a criterion. It can adjust the cruel of the competitive mechanism between corners adaptively. And it can reach a good matching result by combining with singular value decomposition. Experiments show that, this algorithm does not affect the corner matching speed, and the matching precision of corner points have greatly improved.