摘要
针对局部立体匹配算法匹配精度较低问题,提出一种改进Census变换与特征融合的立体匹配算法。首先,使用变换窗口的邻域像素信息代替中心像素,解决传统Census变换过度依赖窗口中心像素问题;其次,引入图像的颜色信息与梯度信息构建融合代价计算函数,提高初始匹配代价的可靠性;为建立邻域像素间联系,引入单向动态规划思想进行代价聚合;最后,提出一种基于八方向的视差空洞填充方法对视差图进行优化。实验结果表明:该算法在Middlebury数据集上的平均误匹配率为3.77%,优于其他改进Census变换方法,具有较高的匹配精度。
Aiming at the problem of low matching precision of local stereo matching algorithm,a stereo matching algorithm based on improved Census transform and feature fusion is proposed.Firstly,the neighborhood pixel information of transformation window is used to replace the central pixel,which solves the problem of traditional Census transform being overly dependent on the center pixel of the window.Then,the color information and gradient information of the image are introduced to construct a fusion cost calculation function to improve the reliability of the initial matching cost.In order to establish the connection between domain pixels,the idea of one-way dynamic programming is introduced for cost aggregation.Finally,a parallax hole filling method based on eight directions is proposed to optimize the disparity map.Experimental results show that average mis-matching rate of the proposed algorithm is 3.77%on the Middlebury dataset,which is superior to other improved Census transform methods and has higher matching precision.
作者
张释如
魏晓艳
ZHANG Shiru;WEI Xiaoyan(College of Communication and Information Engineering,Xi’an University of Science and Technology,Xi’an 710054,China)
出处
《传感器与微系统》
CSCD
北大核心
2024年第2期130-133,共4页
Transducer and Microsystem Technologies
基金
陕西省技术创新引导专项基金资助项目(2020TG—005)。
