非参量变换在彩色图像立体匹配中的应用研究  

Research on Application of Non-parametric Transform in Color Image Stereo Matching

作　　者：郭龙源[1] 卢阿丽[1] 杨静宇[1] 

机构地区：[1]南京理工大学计算机科学与技术系,南京210094

出　　处：《计算机科学》2008年第11期217-219,共3页Computer Science

基　　金：国防基础研究项目(编号:J1500C002)基金资助

摘　　要：随着彩色摄像机广泛使用,对彩色图像的视差匹配研究越来越普遍。研究显示,使用彩色图像匹配比灰度图像匹配精确度增加20%～25%。根据非参量变换原理,提出了基于非参量变换的彩色图像立体匹配算法,将Rank变换应用到彩色图像的区域匹配方法中,并采用census变换后的汉明码距离作为新的约束条件。实验表明提高了原来方法的抗噪性和匹配准确性。Research of color image stereo matching becomes along with wide use of color camera. It is shown that the precision of matching results are always improved by 20% to 25% when using color information instead of gray value information. Presents color image matching approach based on non-parametric transform. With non-parametric theory, applies rank transform to color image local area matching. Hamming distance between pixels after the census transform is presented as a new constraint. The experiment result indicates that the proposed algorithm is the more precise matching and invariance to certain types of image noise than the ones based on intensity.

关 键 词：彩色图像 非参量变换 Rank变换 立体匹配 鲁棒性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN948.41[自动化与计算机技术—计算机科学与技术]