基于局部灰度分布特征的自适应立体匹配算法OA

Targeting at the problem that the traditional AD-Census algorithm has low matching accuracy at the edge and smooth area of the image,and it is difficult to make full use of the local grayscale distribution information of the image,an adaptive stereoscopic matc-hing algorithm based on the local grayscale distribution feature is proposed.Firstly,in the cost calculation stage,the local grayscale dis-tribution feature information of the input image is calculated,the adaptive window of the census transform is selected according to the neighborhood grayscale distribution feature of the pixels,and the adaptive setting of the fusion weight of AD cost and census cost is set.Secondly,an aggregation method based on the improved local grayscale distribution feature is constructed by the fusion of the guide filter and the dynamic programming method.Finally,the final parallax results are obtained after parallax calculation and optimization.The ex-perimental results show that the average mismatch rate of the proposed algorithm in the unoccluded area and the whole area is 3.96%and 7.78%,respectively,which has high accuracy and applicability,and has good robustness to noise.