基于自适应权重的改进Census立体匹配算法

doi:10.19678/j.issn.1000-3428.0057262

摘要/Abstract

摘要： 针对传统Census算法对噪声敏感且在弱纹理区域匹配精度低的不足，提出一种基于自适应权重的改进算法。在代价计算阶段，通过空间相似度加权计算得到参考像素值，设定阈值限定参考值与中心点像素的差异，使算法能够判断中心点是否发生突变并自适应选择中心参考像素值。在代价聚合阶段，引入多尺度聚合策略，将引导滤波作为代价聚合核函数，加入正则化约束保持代价聚合时尺度间的一致性。在视差计算阶段，通过胜者通吃法得到初始视差图。在视差优化阶段，对初始视差图做误匹配点检测及左右一致性检测，并对遮挡区域进行像素填充得到最终的视差图。基于Middlebury标准图的实验结果表明，该算法平均误匹配率为5.81%，对比于传统Census算法抗干扰性提升显著，并能在平均误匹配率表现上达到主流经典算法的性能水准。

关键词: 机器视觉, 立体匹配, Census变换, 自适应权重, 多尺度聚合

Abstract: In view of the noise sensitivity and low matching accuracy of traditional Census algorithms in weak texture region,this paper proposes an improved algorithm based on adaptive weight.At the stage of cost calculation,the reference pixel value is obtained by using the weighted calculation of spatial similarity,and the threshold is set to limit the difference between the reference value and the central point pixel,so that the proposed algorithm can judge whether the central point is mutated or not and select the central reference pixel value adaptively.At the stage of cost aggregation,a multi-scale aggregation strategy is introduced,guided filtering is adopted as the cost aggregation kernel function,and regularization constraints are added to maintain the consistency between scales during cost aggregation.At the stage of disparity calculation,the initial disparity map is obtained by using the Winner-Takes-All(WTA) method.During disparity optimization,mismatching point detection and left-right consistency detection are carried out for the initial disparity map,and pixel filling is carried out for the occluded area to obtain the final disparity map.The experimental results tested based on Middlebury benchmark demonstrate that the average mismatching rate of the proposed algorithm is 5.81%,which reaches the performance standard of mainsteam classical algorithms.The algorithm brings the performance level with the traditional Census algorithm while significantly improving the anti-noise performance.

Key words: machine vision, stereo matching, Census transform, adaptive weight, multi-scale aggregation

中图分类号:

TP391.4

黄彬, 胡立坤, 张宇. 基于自适应权重的改进Census立体匹配算法[J]. 计算机工程, 2021, 47(5): 189-196.

HUANG Bin, HU Likun, ZHANG Yu. Improved Census Stereo Matching Algorithm Based on Adaptive Weight[J]. Computer Engineering, 2021, 47(5): 189-196.

http://www.ecice06.com/CN/Y2021/V47/I5/189

参考文献

[1] SHI Chenbo,WANG Guilin,YIN Xuanwu,et al.High-accuracy stereo matching based on adaptive ground control points[J].IEEE Transactions on Image Processing,2015,24(4):1412-1423.
[2] SCHARSTEIN D,SZELISKI R.A taxonomy and evaluation of dense two-frame stereo correspondence algorithms[J].International Journal of Computer Vision,2002,47(1/2/3):7-42.
[3] VEKSLER O.Stereo correspondence by dynamic programming on a tree[C]//Proceedings of 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2005:384-390.
[4] FELZENSZWALB P F,HUTTENLOCHER D P.Efficient belief propagation for early vision[J].International Journal of Computer Vision,2006,70(1):41-54.
[5] KOLMOGOROV V,ZABIH R.Computing visual correspondence with occlusions using graph cuts[C]//Proceedings of the 8th IEEE International Conference on Computer Vision.Washington D.C.,USA:IEEE Press,2001:508-515.
[6] ZHANG Yifei,LI Xinfu,TIAN Xuedong.SAD stereo matching algorithm combining edge features[J].Computer Engineering,2020,46(4):236-240,246.(in Chinese)张一飞,李新福,田学东.融合边缘特征的SAD立体匹配算法[J].计算机工程,2020,46(4):236-240,246.
[7] MEI Xing,SUN Xun,ZHOU Mingcai,et al.On building an accurate stereo matching system on graphics hardware[C]//Proceedings of 2011 IEEE International Conference on Computer Vision.Washington D.C.,USA:IEEE Press,2011:467-474.
[8] ZABIH R,WOODFILL J.Non-parametric local transforms for computing visual correspondence[C]//Proceedings of the 3rd European Conference on Computer Vision.[S.l.]:DBLP,1994:151-158.
[9] HIRSCHMULLER H,SCHARSTEIN D.Evaluation of stereo matching costs on images with radiometric differences[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2009,31(9):1582-1599.
[10] YOON K J,KWEON I S.Adaptive support-weight approach for correspondence search[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2006,28(4):650-656.
[11] HE Kaiming,SUN Jian,TANG Xiaoou.Guided image filtering[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2013,35(6):1397-1409.
[12] YANG Qingxiong.A non-local cost aggregation method for stereo matching[C]//Proceedings of 2012 IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2012:1402-1409.
[13] ZHANG Kang,FANG Yuqiang,MIN Dongbo,et al.Cross-scale cost aggregation for stereo matching[EB/OL].[2020-01-02].https://arxiv.org/pdf/1403.0316.pdf.
[14] MA Xuan,ZHU Shiqiang,LAI Xiaobo.An image matching algorithm based on improved Rank transform[J].Sensors and Microsystems,2011,30(11):128-130,134.(in Chinese)马璇,朱世强,赖小波.一种基于改进Rank变换的图像匹配算法[J].传感器与微系统,2011,30(11):128-130,134.
[15] LIU Jianguo,YU Li,LIU Sijian,et al.Stereo matching algorithm based on the improved Census transform and multi-scale space[J].Journal of South China University of Technology(Natural Science Edition),2017,45(12):43-49.(in Chinese)刘建国,俞力,柳思健,等.基于改进Census变换和多尺度空间的立体匹配算法[J].华南理工大学学报(自然科学版),2017,45(12):43-49.
[16] NALPANTIDIS L,GASTERATOS A.Stereo vision for robotic applications in the presence of non-ideal lighting conditions[J].Image and Vision Computing,2010,28(6):940-951.
[17] MA Li,LI Jingjiao,MA Ji,et al.A modified census transform based on the neighborhood information for stereo matching algorithm[C]//Proceedings of the 7th IEEE International Conference on Image and Graphics.Washington D.C.,USA:IEEE Press,2013:533-538.
[18] HUMENBERGER M,ZINNER C,WEBER M,et al.A fast stereo matching algorithm suitable for embedded real-time systems[J].Computer Vision and Image Understanding,2010,114(11):1180-1202.

选择文件类型/文献管理软件名称

选择包含的内容