摘要
本文在推导并简化会聚式体视模型视差公式的基础上 ,基于充分的生理依据 ,选取多尺度高斯微分滤波器滤波向量为匹配元 ,采用随机松弛方法的思想 ,设计了一个分层优化三维重建算法。算法利用双向匹配发现存在不利因素区域 ,随后对目标函数各项合理加权 ,从而改善了在该区域的重建效果。该算法能有效地直接产生密集的视差场 ,经实验取得了良好的效果。
In this paper,a disparity formula of the concentrated stereopsis model is deduced and simplified.Based on ample physiological evidences,the response vectors of a bank of multi scale different ial filters are chosen as matching primitives,and an efficient multilevel relaxa tion computational framework for three dimensional reconstruction is developed. With this algorithm,dual matching technique is used to find areas where liabilit y exists,and the cost terms of target function are measured reasonably,thus t he effect of reconstruction of the areas is ameliorated.The proposed algorithm can yield dense disparity field directly and efficiently,and a good result is ob tained through experiments.
出处
《计算机工程与科学》
CSCD
2000年第3期28-31,40,共5页
Computer Engineering & Science
关键词
计算机视觉
会聚式体视模型
三维表面重建
stereo vision
disparity
stereo matching
matchi ng primitive
three dimensional reconstruction
simulated annealing