首页 | 本学科首页   官方微博 | 高级检索  
     检索      

随机立体匹配
引用本文:张永生.随机立体匹配[J].测绘学院学报,1991(2).
作者姓名:张永生
基金项目:国家模式识别开放研究实验室基金
摘    要:本文研究一种实现立体匹配的随机优化方法。与传统的基于面积的相关和基于特征的匹配有所不同,本方法能直接从一对立体图像中提取出稠密的视差值,而根本无须内插。首先,运用正则化理论把立体匹配问题转化为求视差时对两种彼此竞争约束的满足。这些约束用以规定一个动态系统的能量,而该系统的稳态则是通过一种所谓的微正则模拟退火达到的。其次,引入立体图像的Laplacian金字塔结构则有效地实现了由粗到精的分层控制策略。本方法适合于在SIMD体系结构的计算机上并行实现。

关 键 词:立体视觉  立体匹配  模拟退火

STOCHASTIC STEREO MATCHING
Zhang Yongsheng.STOCHASTIC STEREO MATCHING[J].Journal of Institute of Surveying and Mapping,1991(2).
Authors:Zhang Yongsheng
Institution:Zhang Yongsheng
Abstract:A stochastic optimization approach to stereo matching is presented. Unlike conventional area-besed correlation and feature-based matching, the approach provides a dense array of disparities, eliminating the need for interpolaton. First, the stereo matching using regularization is formulated as a problem of finding a disparity map that satisfies two competing constraints are interpreted as specifying the potential energy of a system of oscillators. Ground states of the system are approximated by the called microcanonieal simulated annealing. Secondly, the approach uses a hierarchical, coarse-to-fine strategy employing laplacian pyramids of the stereo images. It is inherently massively parallel, and maps well to SIMD parallel architeture.
Keywords:Stereo Vision  Stereo Matching  Simulated Annealing
本文献已被 CNKI 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号