阈值自适应的极线距离变换算法及其在立体匹配中的应用  

Adaptive Threshold Epipolar Distance Transformation Method with Applications to Stereo Matching

作　　者：王瑞 杨朝旭[1,2,3] 荣海军 张涛[1,2,3] 刘馨媛 WANG Rui;YANG Zhaoxu;RONG Haijun;ZHANG Tao;LIU Xinyuan(State Key Laboratory for Strength and Vibration of Mechanical Structures,Xi’an Jiaotong University,Xi’an 710049,China;Shaanxi Key Laboratory of Environment and Control for Flight Vehicle,Xi’an Jiaotong University,Xi’an 710049,China;School of Aerospace,Xi’an Jiaotong University,Xi’an 710049,China)

机构地区：[1]西安交通大学机械结构强度与振动国家重点实验室,西安710049 [2]西安交通大学陕西省先进飞行器服役环境与控制重点实验室,西安710049 [3]西安交通大学航天航空学院,西安710049

出　　处：《西安交通大学学报》2021年第7期132-140,共9页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(12002254,61976172)。

摘　　要：为了提升图像在不重叠区域和弱纹理区域的匹配率,提出一种基于图像局部纹理信息的阈值自适应的极线距离变换(AT-EDT)算法。选取左右视图对应视差范围内的点,将对应点在特定窗口中的灰度信息作为参考,去除图像不重叠区域;针对在极线距离变换(EDT)中使用固定阈值会产生二次弱纹理和重复纹理的现象,将局部标准差与灰度梯度进行融合来表示图像局部纹理信息,并由此信息获得图像中每个像素对应的距离变换阈值;利用AT-EDT算法对图像进行变换,并通过立体匹配得到最终视差图。实验结果表明:利用AT-EDT算法可有效去除图像的不重叠区域并显著提升弱纹理区域的像素可分性,同时避免了EDT算法导致的二次弱纹理和重复纹理现象;对变换后的图像进行立体匹配,可有效提升视差图的深度连续性;与EDT后的图像立体匹配得到的视差图相比,AT-EDT的匹配率至少提升了20%。To improve the matching rate of images in non-overlapping regions and weak texture regions,an adaptive threshold epipolar distance transformation(AT-EDT)method related to image local texture information is proposed.The points in the disparity range corresponding to the left and right views are selected,and the gray information of the corresponding point in a specific window is taken as a reference to remove the non-overlapping area of the image.Aiming at the secondary weak texture and repeated texture that will appear as a fixed threshold is used in epipolar distance transformation(EDT),the local standard deviation and gray gradient are fused to represent the local texture information of the image,and the corresponding distance transformation threshold of each pixel in the image is obtained from this information.The AT-EDT method is chosen to transform the image,and the disparity map is obtained by stereo matching.Experimental results show that the AT-EDT method can effectively remove the non-overlapping regions of the image and significantly improve the pixel separability of the weak texture regions while the secondary weak texture and repetitive texture caused by the EDT algorithm are avoided.The stereo matching of the transformed image can improve the depth continuity of disparity map.Compared with the disparity map obtained by the stereo matching of the image after EDT,the disparity map matching rate is heightened by at least 20%.

关 键 词：立体匹配 弱纹理 阈值自适应 极线距离变换 图像重叠区域分割 

分 类 号：TP302.7[自动化与计算机技术—计算机系统结构]