基于跨尺度代价聚合的改进立体匹配算法  

Improved Stereo Matching Algorithm Based on Cross-scale Cost Aggregation

作　　者：汤春明[1] 蒋昂 

机构地区：[1]天津工业大学电子与信息工程学院,天津300387

出　　处：《计算机工程》2016年第11期272-276,280,共6页Computer Engineering

基　　金：天津市第三批"三年千人"计划项目(62014511);天津工业大学高层次人才及教师引进计划项目(030367)

摘　　要：针对现有的立体匹配算法在精度和速度上不可兼得的现状,提出一种改进的跨尺度代价聚合立体匹配算法。通过强度和梯度算法计算匹配代价,利用引导滤波聚合匹配代价,采用跨尺度模型聚合各尺度的匹配代价,使用补丁匹配近似算法代替传统胜者为王算法求取初始视差,并在视差细化阶段采用快速加权中值滤波进行后续处理。实验结果表明,该算法能在快速获得视差图的同时提高匹配精度。An improved algorithm is proposed based on cross-scale cost aggregation for stereo matching because the existing stereo matching algorithms cannot take both accuracy and speed into account.Firstly the matching cost volume is computed by the intensity and gradient algorithm.Then the matching cost volume is aggregated by guide filtering and the matching cost volume of different scales is aggregated by the cross-scale model.The patch matching approximation algorithm is used instead of the traditional Winner Taker All（WTA）algorithm to calculate the initial parallax,and the weighted median filtering is adopted for subsequent processing in parallax elaboration.Experimental results show that the algorithm can rapidly obtain the disparity map and also improve the matching accuracy.

关 键 词：立体匹配算法 跨尺度模型 代价聚合 引导滤波 快速加权中值滤波 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]