基于多种几何约束的误匹配剔除算法
A Mismatch Removing Method Based on the Combination of Multiple Geometric Constraints
投稿时间:2019-06-06  修订日期:2019-06-06
DOI:
中文关键词: 图像匹配  误匹配剔除  几何约束  组合策略
英文关键词: image matching  mismatch removing  geometric constraint  combinatorial strategy
基金项目:国家自然科学基金面上项目(61773409);中央高校基本科研业务费专项资金项目(CZY17011)
作者单位E-mail
刘李漫 中南民族大学生物医学工程学院 limanliu@mail.scuec.edu.cn 
谭龙雨 中南民族大学生物医学工程学院  
刘佳 中南民族大学生物医学工程学院  
摘要点击次数: 305
全文下载次数: 0
中文摘要:
      图像匹配是数字图像处理的基本问题。图像匹配精确度的高低,直接影响到许多其它应用的效果。针对现有特征点匹配算法包含较多错误匹配的问题,本文以尺度不变特征SIFT匹配算法为例,提出了一种基于多种几何约束的误匹配剔除算法。该方法利用匹配点对之间多种几何约束的组合来识别并剔除错误的匹配,具体包括方向角算法、邻域信息算法、点线距离算法和局部相似三角形算法。实验结果证明,这一基于组合策略的方法能有效地剔除错误匹配,并且极大限度地保留正确匹配。
英文摘要:
      Image matching is the fundamental problem of digital image processing. The precision of matches greatly affects the performance of other applications. However, current matching methods include many false matches in the result. To tackle this, in this paper we take the SIFT matching method as an example and propose a mismatch removing method based on the combination of multiple geometric constraints. This method combines multiple geometric constraints between match pairs to identify and then discard false matches. Specifically, the proposed method combines four sub-algorithms including the direction angle algorithm, the neighboring information algorithm, the point-line algorithm and the local similar triangle algorithm. Experiments show that the combinatorial strategy can effectively remove false matches while preserving the most correct matches.
View Fulltext   查看/发表评论  下载PDF阅读器
关闭