基于颜色变化信息的局部立体匹配算法

Local Stereo Matching Algorithm Based on Color Information

主要针对传统的局部匹配算法中遮挡区域、重复区域和不连续区域的精度不高问题,提出了一种基于遮挡信息和颜色变化约束的局部匹配算法。首先,提出一种基于颜色变化约束的区域生长算法来提高权重因子的准确性,有效提高不连续区域、重复区域和低纹理区域的匹配精度;接着针对遮挡区域难匹配问题,提出通过降低遮挡像素点的权重影响来提高支持聚合的思想;最后,结合生成的区域和遮挡点对权重因子进行分类,最终生成视差图。实验结果表明,该算法能有效提高重复区域、边缘区域和遮挡区域的匹配精度,与当前主流算法具有可比性。

For local adaptive algorithm always has lower accuracy in occluded areas, repetitive texture areas and depth discontinuities, a novel local stereo matching method based on occlusion information and color changing constraint is presented in this paper. Firstly ,a region growing algorithm based on color changing constraint is proposed to improve the matching accuracy of depth discontinuities, repetitive and low texture areas effectively. On the other side, support aggregation idea by the way of reducing occluded pixels＇ s weight influence is presented to solve matching problem of occluded pixels. Finally, a dense disparity map is obtained by growing region and occluded pixels. Experimental results show this algorithm can effectively improve the matching accuracy in repetitve and depth discontinuities as also as low-textured regions and occluded areas,and get a better stereo matching quality which is comparable with other main stream stereo methods.