Research based on improved AD-Census stereo matching algorithm

Abstract

Aiming at the problem that the synthesis of AD and Census matching costs of AD-Census transformation cannot reflect the weight and central pixel dependence of pixels in the image area, an improved AD-Census stereo matching algorithm is proposed. The algorithm selects the reference pixel values by assigning the distance similarity of the pixels in the support window, and sets the threshold range based on the center pixel difference, and then calculates the cost of AD transformation cost and Census transformation cost composite weight. By using a semi-global aggregation algorithm based on dynamic programming, SGM (Semi-global matching) algorithm, the matching cost information is fused. The results of the Middlebury stereo matching evaluation platform show that compared with the binocular stereo matching performance of the improved AD-Census transform compared with the binocular stereo matching performance based on AD-Census transform, the parallax error of pixels is significantly optimized, and by adding noise interference, it can be seen that the anti-interference performance of the improved AD-Census algorithm is improved.