基于互信息和单应性原理的图像自动配准研究

Research on Image Automatic Registration Based on Mutual Information and Homography Theory

基于图像特征的配准是图像配准的一个主要研究方向。传统的基于特征的配准方法一般都存在着计算量大、复杂度高以及配准精度低等问题。该文提出了一种基于互信息和单应性原理的图像自动配准方法。该方法首先利用harris算子提取角点,然后通过互信息粗匹配和RANSAC精匹配得到一对一的匹配角点对,最后根据匹配角点对计算出对应的单应矩阵来实现图像配准。为相互间满足刚体变换关系的图像之间的配准提供了一种新的思路。

The image registration based on the features of the image is a major research direction.It is generally that there are many problems in traditional feature registration,e.g,a large amount of computation,high complexity and low accuracy.So,an image automatic registration method based on mutual information and homography theory is given out.First,the feature points are detected using Harris feature detector,then the one to one matching points could be obtained through mutual information rough match and RANSAC...