加入结构约束的半全局立体匹配方法

将图像分割得到的场景结构信息作为一种软约束加入到半全局立体匹配的框架中,以克服原算法存在的传播路径不完全的问题,同时结构约束还被用来解决遮挡和错误匹配区域深度选择的问题.实验结果表明,该算法显著地提高了匹配精度,很好地解决了传播路径不完全引起的深度污染问题,并且为遮挡和错误匹配区域提供了可靠的深度估算.同时,算法并不破坏原算法的基本框架,保留了其高效的特点以及使用单指令多数据流技术进行加速的能力,具有广泛的实用价值.

SEMI-GLOBAL STEREO MATCHING COMBINED WITH STRUCTURE CONSTRAINT

This paper presents an improved Semi-global stereo matching algorithm framework combined with structure information of scene as a soft constraint which is extracted from image segmentation.Structure constraint is used to solve the problem of incomplete path in the former algorithm and provide estimates to occlusion and mismatch area depth selection.Experiments demonstrate that the match accuracy is improved significantly by solving depth pollution caused by incomplete path,and by providing reliable depth es...