Improved AD-Census stereo matching algorithm based on fusion gradient information

In view of the drawback that the stereo matching algorithm cannot simultaneously improve the accuracy and speed, an improved stereo matching algorithm based on AD-Census fusion of gradient information is proposed. The Census transformation based on parameters is improved to the Census transformation of the eight domains within the central pixel point window, and combined with the Absolute Difference AD algorithm for local matching. This greatly shortens the time consumption of the cost calculation stage. To simultaneously improve the matching accuracy, the gradient information of pixels is integrated into the cost matching process. Experimental results show that the cost calculation time of the algorithm on images with simple pixels is stable within 0.4 seconds, and the mismatch rate on the Middlebury2.0 test platform is reduced by about 5% compared with the unoptimized algorithm.