Stereo matching is the focus of research in the field of computer vision. Most of the stereo matching algorithm take mathematical calculation as digital information, and lack the connection with the actual human visual characteristics. We combine the concentric antagonistic receptive field of the human eye and the HSI colour space model to improve the weight calculation of the original adaptive support weight (ASW), and optimize the parallax by means of left-right consistency check and median filtering. The test results of several groups of international standard images in VS2010 platform show that, this method has higher matching accuracy than that of the original ASW algorithm in the low texture regions and depth discontinuities area. According to the difference of the test images, the improvement ranges from 10% to 20%, and the overall matching accuracy is similar to that of the mainstream local matching method in recent years.