基于图像分割的稠密立体匹配算法 | |
Alternative Title | Dense Stereo Matching Based on Image Segmentation |
马瑞浩1,2,3; 朱枫1,3![]() ![]() ![]() | |
Department | 光电信息技术研究室 |
Source Publication | 光学学报
![]() |
ISSN | 0253-2239 |
2019 | |
Volume | 39Issue:3Pages:1-9 |
Indexed By | EI ; CSCD |
EI Accession number | 20192006931161 |
CSCD ID | CSCD:6461045 |
Contribution Rank | 1 |
Funding Organization | 国家自然科学基金(U1713216) |
Keyword | 机器视觉 立体匹配算法 匹配代价计算方法融合 匹配代价聚合 十字交叉自适应窗口加权中值滤波 |
Abstract | 针对立体匹配中匹配代价难以选择和物体边缘误匹配的问题,提出一种基于图像分割的稠密立体匹配算法。根据目前立体匹配代价计算方式的特点,提出将灰度加梯度算法与零均归一化互相关算法ZNCC(Zero-mean Normalized Cross-Correlation algorithm)相融合的方法,兼顾这两种算法各自的优势。利用SLIC(Simple Liner Iterative Cluster)算法对图像进行分割,根据图像像素是否在超像素边缘,赋予两种算法不同的权值,提高匹配代价的准确率。在匹配代价聚合阶段,提出基于视差图和超像素更新匹配代价的方法。在视差后处理阶段,采取左右一致性检验LRC (Left Right Check)、孔洞填充和十字交叉自适应窗口加权中值滤波的方法减小视差图的误匹配率。在Middlebury数据集的4组图像上进行测试,测试结果表明平均误匹配率为4.99%。 |
Other Abstract | Aiming at the problems of choosing matching cost computation methods and mismatching at the edge of objects in stereo matching, a dense stereo matching algorithm based on image segmentation is proposed. According to the features of the present matching cost computation methods, an algorithm which combines the advantages of the intensity and gradient algorithm and ZNCC (Zero-mean Normalized Cross-Correlation algorithm) is proposed. The SLIC(Simple Liner Iterative Cluster) algorithm is used for image segmentation. And depending on whether the image pixel is at the super pixel edge, different weights are assigned to the two algorithms in order to improve the accuracy of the matching cost. At the matching cost aggregation stage, a method based disparity and super pixel is proposed to generate new matching cost. At the disparity post-processing stage, Left Right Check(LRC)、Hole Filling and Cross Adaptive Window Weighted Median Filtering are used to reduce the error matching rate of the disparity. Performance evaluation experiments on 4 Middlebury stereo pairs demonstrate that the proposed algorithm achieves an average error matching rate of 4.99%. |
Language | 中文 |
Citation statistics | |
Document Type | 期刊论文 |
Identifier | http://ir.sia.cn/handle/173321/23415 |
Collection | 光电信息技术研究室 |
Corresponding Author | 朱枫 |
Affiliation | 1.中国科学院沈阳自动化研究所 2.东北大学信息科学与工程学院 3.中国科学院光电信息处理重点实验室 |
Recommended Citation GB/T 7714 | 马瑞浩,朱枫,吴清潇,等. 基于图像分割的稠密立体匹配算法[J]. 光学学报,2019,39(3):1-9. |
APA | 马瑞浩,朱枫,吴清潇,鲁荣荣,&魏景阳.(2019).基于图像分割的稠密立体匹配算法.光学学报,39(3),1-9. |
MLA | 马瑞浩,et al."基于图像分割的稠密立体匹配算法".光学学报 39.3(2019):1-9. |
Files in This Item: | ||||||
File Name/Size | DocType | Version | Access | License | ||
基于图像分割的稠密立体匹配算法.pdf(607KB) | 期刊论文 | 作者接受稿 | 开放获取 | CC BY-NC-SA | View Application Full Text |
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment