像素扩展自适应窗口立体匹配算法  

Stereo matching algorithm for adaptive window based on pixel expansion

作　　者：门宇博 张国印[1] 门朝光[1] 李香[1] 马宁[1] MEN Yubo;ZHANG Guoyin;MEN Chaoguang;LI Xiang;MA Ning(College of Computer Science and Technology, Harbin Engineering University, Harbin 150001, China)

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,黑龙江哈尔滨150001

出　　处：《哈尔滨工程大学学报》2018年第3期547-553,共7页Journal of Harbin Engineering University

基　　金：国家自然科学基金项目(61672181)

摘　　要：针对局部立体匹配方法中的视差图前景膨胀问题,依据格式塔视觉理论,采用基于像素扩展的自适应窗口立体匹配方法减少视差图中前景膨胀现象的产生。该方法首先通过扩展约束条件建立待匹配像素点窗口模型;然后通过待匹配点对的窗口相对重合率进行预匹配判定,在视差搜索范围内逐点筛选出满足匹配条件的窗口区域;最后利用窗口规则化的互相关系数方法以及平滑视差平面分割获得最终视差值。实验结果表明:该算法可以获得较准确的稠密视差图,提高立体像对中深度不连续区域和遮挡区域的匹配精度,在Bicycle2、Classroom2、Hoops和Staircase图像中平均误匹配率为14.9%。Aiming at the foreground expansion in the disparity map of the local stereo matching method,this study uses the adaptive window stereo matching method based on pixel expansion to reduce the foreground expansion in the disparity map according to the Gestalt vision theory.First,the method establishes the window model for the pixel to be matched by the extending constraints;second,the pre-matching judgment is conducted using a relative coincidence rate of the window of the point pairs to be matched,and the window that meets the matching condition is selected point-by-point in the disparity searching range.Finally,the final disparity result is calculated through the normalized cross correlation of window regularization and the smooth disparity plane segmentation.The experimental results show that the proposed algorithm can obtain a more accurate dense disparity map and improve the matching accuracy of the depth discontinuities and occlusion regions.The average mismatching rate in Bicycle2,Classroom2,Hoops,and Staircase is14.9%.

关 键 词：计算机视觉 立体匹配 代价聚合 前景膨胀 自适应窗口 视差计算 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]