摘要
图像的立体匹配一直是立体视觉的研究重点。首先简要介绍了立体匹配方法及其分类,归纳了立体匹配中的各种约束条件和相似性测度函数;然后总结了局部匹配算法和全局匹配算法的特点,并结合对象的三维重建问题重点分析了全局匹配算法中的动态规划算法、图割法和置信度传播算法;最后对立体匹配研究面临的主要问题给出了一些建议。
Image stereo matching is the most important problem in stereo vision. Stereo matching and its categories are briefly introduced. After the matching constraints and measurements of similarity are concluded, some local matching algorithms and global matching algorithms are also analyzed. Considering object 3D reconstruction problem, three typical algorithms, such as dynamic programming algorithm, graph cuts algorithm and belief propagation algorithm are addressed in detail. Finally, some major problems in stereo matching are pointed out and some suggestions are given.
出处
《测控技术》
CSCD
北大核心
2009年第8期1-5,10,共6页
Measurement & Control Technology
基金
国家高技术研究发展计划863(2007AA704310)
关键词
立体匹配
匹配约束
相似性测度
stereo matching
matching constraint
measurement of similarity
