基于插值优化的动态规划立体匹配算法
首发时间:2015-11-06
摘要:本文提出了一种基于插值优化的动态规划立体匹配算法。该算法使用左右图分别作为主视图,计算匹配代价后迭代计算对应的视差图,然后使用插值优化修正误匹配点,得到精确的视差图。其中,计算匹配代价同时考虑像素相似度和纹理相似度,迭代过程使用动态规划算法,兼顾实时性和精确度,并针对动态规划算法的条纹特性以及同一扫描线上的精确度不对等,使用插值优化算法提取左右视差图的高精度区域合成最终的视差图。实验结果表明,同传统动态规划算法相比,基于插值优化的动态规划立体匹配算法在遮挡区域表现优异,并且依然保有实时性。
For information in English, please click here
An Interpolation Refinement Dynamic Programming Algorithm for Stereo Matching
Abstract:This paper presents an interpolation refinement Dynamic Programming algorithm for stereo matching. The program uses left and right image as basic image, respectively calculates the matching cost and iterate left and right disparity map. Interpolation refinement is used to fix the error matching point afterwards. Among them, the matching cost is calculated by AD-census,which balances the similarity of color and the texture of the matching point. Dynamic programming is used for the cost calculation step, as a result of its high precision and real-time capabilities. Due to the characteristics of dynamic programming: horizontal streaks and uneven of precision on one scanning line, interpolation refinement is proposed to obtain the high precision region to redraw the disparity map. Experimental results show that compared to the traditinal dynamic programming, interpolation refinement dynamic programming has better performance and real-time capabilities.
Keywords: interppolation refinement AD-census dynamic programming stereo matching
基金:
论文图表:
引用
No.4658779109489214****
同行评议
共计0人参与
勘误表
基于插值优化的动态规划立体匹配算法
评论
全部评论0/1000