| 基于改进 Census 变换的图像立体匹配算法研究 |
| |
| 引用本文： | 王丹,杨家琪,杨谢柳.基于改进 Census 变换的图像立体匹配算法研究[J].机械与电子,2021,39(5):62-67. |
| |
| 作者姓名： | 王丹 杨家琪 杨谢柳 |
| |
| 作者单位： | 沈阳建筑大学机械工程学院,辽宁 沈阳 １１０１６８ |
| |
| 摘 要： | 针对传统的 Census 区域匹配算法过分依赖窗口中心像素信息,导致算法受到噪声干扰时匹配精度降低的问题,提出一种基于改进 Census 变换的匹配算法.采用局部像素反差值为中心像素选择的评判标准,对传统的 Census 变换进行改进,增强了窗口像素信息的利用,提高了算法对像素值突变的适应性,使算法有更好的鲁棒性;代价聚合阶段采用引导图滤波算法并结合多尺度聚合模型,增强平坦区域像素间的区分度;采用 win-take-all 算法选取最优视差值,完成视差计算;采用区域投票策略和中值滤波算法完成视差精化.利用该改进算法对 Middlebury 平台提供的标准图像进行实验,实验结果表明该算法较传统Census 算法有较好的抗噪能力和立体匹配精度.
|
| 关 键 词： | 机器视觉 立体匹配 Census 变换 多尺度空间 视差 |
Research on Image Stereo Matching Algorithm Based on Improved Census Transform |
| |
| WANG Dan,YANG Jiaqi,YANG Xieliu.Research on Image Stereo Matching Algorithm Based on Improved Census Transform[J].Machinery & Electronics,2021,39(5):62-67. |
| |
| Authors: | WANG Dan YANG Jiaqi YANG Xieliu |
| |
| Affiliation: | ( College of Mechanical Engineering ,Shenyang Jianzhu University , Shenyang １１０１６８ , China |
| |
| Abstract: | Aiming at the problem that the traditional Census area matching algorithm relies too much on the pixel information of the center of the window,resulting in the reduction of matching accuracy when the algorithm is interfered by noise,a matching algorithm based on improved Census transform is proposed．Using the local pixel contrast value as the criterion for selecting the central pixel , the traditional Census transform is improved , the use of window pixel information is enhanced , th ealgorithm＇s adaptability? to pixel value mutations is improved ,and the algorithm is more robust ; In the cost aggregation stage , the guide graph filtering algorithm and the multi- scale aggregation model are used to enhance the discrimination between pixels in flat areas ; the win-take-all algorithm is used to select the optimal parallax value to complete the parallax calculation ; the regional voting strategy and median filtering are used．Thealgorithmcompletestheparallaxrefinement．The improved algorithm was used to experiment on the standard images provided by the Middlebury platform．The experimental results show that the algorithm has better noise immunity and stereo matching accuracy than the traditional Census algorithm． |
| |
| Keywords: | machine vision stereo matching Census transform multi-scale space parallax |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《机械与电子》浏览原始摘要信息 |
|
点击此处可从《机械与电子》下载全文 |
|
