A fast stereo matching algorithm based on dynamic programming on limited search space is proposed to attain an accurate dense disparity map. The proposed algorithm is built on the base of dynamic programming for stereo matching and it limits the search space by an initial match sequence to reduce the number of vertices to be expanded. The multi-window aggregation strategy based on adaptive weight is proposed to improve the matching accuracy. The brightness gradient is integrated into the smoothness constraints to avoid over-punishment on disparity discontinuity on object boundary. Experimental results show that the matching speed and accuracy of the proposed algorithm are greatly improved, and the algorithm is simple and effective for stereo matching.%为快速而准确地得到稠密视差图,提出了一种基于限制搜索空间的动态规划立体匹配算法。该算法以动态规划立体匹配方法为基础,通过初始匹配序列限制搜索空间以减少搜索变量个数。同时,提出一种基于自适应权重的多窗口累积策略来提高匹配精度,并在平滑性限制中引入亮度梯度以避免在物体边界的视差不连续处产生过度惩罚。实验结果表明,该匹配算法在匹配速度和匹配精度上都有很大的提高,是一种简单有效的立体匹配算法。
展开▼