一种改进的立体像对密集点匹配算法  

A Modified Image Dense Stereo Matching Algorithm

作　　者：王瑞[1] 杨润泽[1] 尹晓春[1] 

机构地区：[1]军械工程学院机械制造教研室,河北石家庄050003

出　　处：《计算机技术与发展》2011年第9期70-72,76,共4页Computer Technology and Development

基　　金：军械工程学院基础部科学研究基金(JCB1011)

摘　　要：文中研究在随机光场照射下基于窗口的密集点匹配算法。针对基于窗口匹配中出现的细节平滑问题,提出了在窗口内对像素点引入权值参数来提高细节部分测量精度的方法。在匹配前对图像进行校正,使得各极线和图像坐标系的横轴平行,在匹配过程中不需要再计算极线方程,大大提高了匹配的效率。根据待匹配图像的灰度信息和摄像机的内外参数信息,通过求解对应窗口内像素点向量夹角的方式进行双目立体匹配。在试验中,完成了棋盘格图像的密集点匹配,并与典型区域匹配算法进行了匹配效率与视差图效果的对比分析,验证了文中算法在匹配效率和视差图效果方面的优越性。It studies the random field irradiation window-based dense point matching algorithm.Matching based on window appears for the details of a smooth problem,a pixel in the window on the introduction of weighting parameters to improve the details of the part of the measurement accuracy.Correct the image before the match,making the polar coordinate system parallel to the horizontal axis,in the matching process does not require re-calculation of polar equation,greatly improving the efficiency of matching.According to the grayscale information and the camera parameter information inside and outside,by solving the corresponding pixel angle within the window between vectors conduct stereo matching.In the study,completed an intensive checkerboard image point matching and compared with the typical area of the matching algorithm efficiency and effectiveness of disparity map,verify the superiority of the algorithm in matching efficiency and effect of disparity map.

关 键 词：双目立体视觉 立体匹配 细节平滑 密集点匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]