A Stereo Matching Algorithm Based on Directional Cost Aggregation
一种基于方向能量聚集的立体匹配算法

A new local stereo matching algorithm is proposed in this paper. The process starts with classifying pixels in the reference image into homogeneous pixels and heterogeneous ones; Then, N directional cost aggregation is applied to heterogeneous pixels. The optimal disparities of all directions are picked out using WTA method, and the disparity who has the maximum statistic is the finally disparity. Homogeneous pixels are matched using shiftable rectangle window for aggregating more pixels. At last, a fast an...