摘要
PatchMatch算法是多视密集匹配中一种常用的算法,但其核心算法以随机数为基础展开,匹配结果存在稳定性过低的问题。本文对PatchMatch多视密集匹配算法进行了改进,在匹配代价、深度信息初始化和深度信息随机搜索中充分应用了点线信息,尽可能避免随机数带来的不确定性。最后本文使用Strecha数据集进行了测试,结果表明,本文算法能有效避免由于随机数导致的匹配结果不稳定的问题,获得了非常稳健的重建结果。
The PatchMatch Algorithm is very common in MultiView Dense Matching,but it’s reconstruction result is not stable due to the random part in the algorithm.In this paper,we improve the PatchMatch Algorithm by Line and Point Information,which is used in the Calculation of Matching Cost,the Initialization of Depth Data and the Random Searching of Depth Data.Finally,we test our algorithm on the Strecha Dataset and the results show that our algorithm can avoid the instability caused by the random and has a much more stable reconstruction result.
作者
颜青松
邓非
曲英杰
陈欣
YAN Qingsong;DENG Fei;QU Yingjie;CHEN Xin(School of Geodesy and Geomatics,Wuhan University,Wuhan 430079,China;Key Laboratory of Urban Land Resources Monitoring and Simulation,Shenzhen 518000,China)
出处
《测绘地理信息》
CSCD
2021年第S01期257-260,共4页
Journal of Geomatics
基金
国土资源部城市土地资源监测与仿真重点实验室开放基金(KF-2018-03-025)
