基于图像分割的改进Census变换立体匹配算法  

Improved census transform for stereo matching algorithm based on image segmentation

作　　者：王书侠 王国新 康辉 WANG Shuxia;WANG Guoxin;KANG Hui(School of Electronics and Information Engineering,Heilongjiang University of Science and Technology,Harbin 150022,China;School of Electrical and Control Engineering,Heilongjiang University of Science and Technology,Harbin 150022,China)

机构地区：[1]黑龙江科技大学电子与信息工程学院,黑龙江哈尔滨150022 [2]黑龙江科技大学电气与控制工程学院,黑龙江哈尔滨150022

出　　处：《高师理科学刊》2021年第4期35-39,共5页Journal of Science of Teachers＇College and University

基　　金：2020年度黑龙江省省属高等学校基本科研业务费科研项目(702/0000070210)。

摘　　要：针对传统Census变换立体匹配算法精度低,提出基于图像分割的改进Census变换立体匹配算法.将Census变换窗口中心像素的灰度值用其邻域像素灰度值的中值代替,从而减少噪声对匹配精度的影响,增强算法的抗干扰能力.采用MeanShift算法对图像进行分割,结合其Census变换结果,以Hamming距离作为匹配代价,求取最佳匹配点,进而求取初始视差.通过左右一致性检验对视差图进行细化,对误匹配点进行替代和对遮挡区域进行填充,生成最终视差图.实验显示,该算法的鲁棒性和匹配精度优于传统算法,并且在深度不连续区域也能够获得精确的视差.In order to improve the accuracy of stereo matching based on typical Census transform,a stereo matching algorithm using improved Census transform based on image segmentation is proposed.Use the median of neighborhood pixels to replace the gray value of central pixel in the transform window,to reduce the influence of noise on the matching accuracy,and enhance the anti-interference ability of the algorithm.Use Mean Shift algorithm for image segmentation,combine with the Census transformation result,use the Hamming distance as the matching cost to calculate the best match point,and then to calculate the initial disparity.The last disparity map is obtained by means of left-right checking,to replace the miss matching point and occlusion filling.Experiment results indicate that this algorithm′s robustness and matching accuracy are better than traditional algorithm.Accurate disparity could be obtained even in the depth-discontinuities regions.

关 键 词：立体匹配 Census变换 图像分割 HAMMING距离 左右一致性检验 

分 类 号：TN911.73[电子电信—通信与信息系统]