摘要
针对当前Census变换过程中对于中心像素过于依赖的问题,提出了一种基于改进的Census立体匹配算法。求取原始Census变换窗口中周围像素的平均值,并用求取到的平均值替换掉原有的中心像素进行匹配代价计算,从而减少原有算法对于中心像素的依赖,改善了中心像素与领域像素的平衡。在公共测试平台上提供的标准测试图像对本文算法进行反复测试,并与其他匹配算法进行比对。根据实验结果可以得知,经本文算法处理后得到的视差图平均误匹配率为6%,相比于原算法显著提高了匹配精确度。
Aiming at the problem of excessive dependence on the center pixel in the current Census transformation process,an improved Census stereo matching algorithm is proposed.The average value of surrounding pixels in the original Census transform window is calculated and used to replace the original center pixel when determining the matching cost,thereby reducing the dependence of the original algorithm on the center pixel and improving the center pixel and the field pixel balance.The standard test images provided on the public test platform are used to repeatedly test the algorithm in this paper and compare it with other matching algorithms.According to the experimental results,the average mismatch rate of the disparity map obtained by this algorithm is 6%,which significantly improves the matching accuracy compared with the original algorithm.
作者
李忠国
吴昊宸
陆军
罗晓钟
王琪
席茜
LI Zhongguo;WU Haochen;LU Jun;LUO Xiaozhong;WANG Qi;XI Qian(College of Mechanical Engineering,Jiangsu University of Science and Technology,Zhenjiang Jiangsu 21200,China;Suzhou Kangli Orthopedic Instrument Co.LTD,Suzhou Jiangsu 215600,China;Zhangjiagang CIMC Sanctum Cryogenic Equipment Co.,LTD,Zhangjiagang Jiangsu 215600,China)
出处
《机械设计与研究》
CSCD
北大核心
2022年第3期27-29,36,共4页
Machine Design And Research
基金
国家重点研发计划资助项目(2018YFC0309100)。
