一种应用于图像配准中的点特征匹配算法  

An Algorithm for Point Pattern Matching Applied to the Registration of Images with Relatively Great Affine Geometric Distortion

作　　者：田伟刚[1] 郭雷[1] 黄雷[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《微电子学与计算机》2008年第3期172-174,181,共4页Microelectronics & Computer

基　　金：国家自然科学基金项目(60175001)

摘　　要：点特征匹配在机器视觉、图像配准等领域中有着重要的应用.针对空间存在较大仿射几何差异的图像中的点特征匹配问题,提出了一种利用马氏距离仿射不变性进行约束的松弛匹配算法,并将该算法应用于遥感图像配准中.实验结果表明,算法可以很好的完成点特征匹配,匹配点对数量充足且具备很高的正确率,从而可以保证图像配准的精度.Point pattern matching is a common problem in machine vison and image registration. To match points from images with relatively great affine geometric distortion, a rdaxation algorithm restricted by the invariance of Mahalanobis distance in affine geometric distortion is proposed and used in the registration of remote sensing image. Experimental result shows that sufficient correct matching-points are accquired by the algorithm and can be applied to registration of images with relatively great affine geometric distortion.

关 键 词：点特征匹配 松弛算法 马氏距离 仿射几何变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]