异形窗口快速立体匹配算法研究  

Research on fast stereo matching algorithm of special-shaped window

作　　者：杨兴梅 黄林海 吴晓松[1,2,3,4] 顾乃庭 YANG Xingmei;HUANG Linhai;WU Xiaosong;GU Naiting(National Key Laboratory of Optical Field Manipulation Science and Technology,Chinese Academy of Sciences,Chengdu 610209,China;The Key Laboratory on Adaptive Optics,Chinese Academy of Sciences,Chengdu 610209,China;Institute of Optics and Electronics,Chinese Academy of Sciences,Chengdu 610209,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院光场调控科学技术全国重点实验室,成都610209 [2]中国科学院自适应光学重点实验室,成都610209 [3]中国科学院光电技术研究所,成都610209 [4]中国科学院大学,北京100049

出　　处：《重庆理工大学学报（自然科学）》2024年第9期158-163,共6页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金项目(12073032,12022308);中国科学院战略性先导科技专项(A类)(XDA25020316)。

摘　　要：针对传统的立体匹配算法需要为图像中的每个像素选择一个合适的支撑窗口,时间成本较大,并且窗口的形状和大小难以确定的问题,提出一种异形窗口的立体块匹配算法。该算法根据灰度相似性和空间邻近性选择匹配窗口,同时对目标图像中搜索窗口内的像素添加阈值范围的约束,减小深度不连续像素和遮挡像素对结果的影响。进行匹配计算,并将计算的视差作为窗口内所有像素的视差,减少逐像素计算视差。利用Middlebury官方数据集中的4组图像对算法的性能进行仿真评估。结果表明:所提算法在视差图精度基本一致的情况下,明显提高了计算速度。The traditional stereo matching algorithms have to choose a suitable support window for each pixel in the image,which is time-consuming and has difficulties in determining the shape and size of the window.To address these issues,this paper proposes a fast stereo matching algorithm of special-shaped window.It selects matching windows based on image gray-scale similarity and spatial proximity,and adds a threshold constraint to the pixels in the search window in the target image to reduce the impact of depth discontinuous regions and occluded regions on the results.Then,the matching calculation is conducted and the calculated parallax is considered to be the parallax of all pixels in the matching window,thus reducing the process of parallax calculation by pixel.The performance of algorithm is verified through numerical simulations using four sets of images from the Middlebury dataset.Our results show the proposed algorithm markedly improves the calculation speed when the accuracy of the disparity map is largely consistent.

关 键 词：机器视觉 立体匹配 异形窗口 遮挡处理 视差填充 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]