一种快速鲁棒区域匹配算法  

A Fast And Robust Are-based Stereo Matching Algorithm

作　　者：刘维[1] 罗桂娥[1] 杨欣荣[1] 

机构地区：[1]中南大学信息科学与工程学院

出　　处：《微计算机信息》2009年第31期184-185,191,共3页Control & Automation

摘　　要：基于区域的匹配算法是立体匹配中常用的方法之一。针对传统的区域匹配算法匹配错误率高,抗噪性低及匹配速度慢等缺点,提出了一种快速鲁棒的区域匹配算法。文中提出的新的相似性测度函数提高了匹配的鲁棒性,同时视差梯度的引入减少了匹配的搜索空间,极大的提高了匹配速度。实验结果表明,本文算法匹配错误率低于SSD算法匹配速度较SSD算法提高了一倍多。The are-based stereo matching algorithm is the commonly used method in stereo matching. Due to the traditional are- based stereo matching algorithm have high matching error rate,sensitive for noise and matching speed low,a fast and robust are-based stereo matching algorithm is proposed. In this paper a new similarity function is employed to improve the robust of matching.At the same time ,the disparity gradient is used to reduce matching search space and improve matching speed largely.The result of experimental prove that the algorithm in this paper have lower matching error rate than SSD algorithm and doubled the matching speed of the SSD algorithm.

关 键 词：立体匹配 视差梯度 鲁棒性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]