摘要： 传统的互信息法直接对大图进行互信息配准，容易出现局部极值并产生误匹配现象，针对该问题，提出一种将互信息与模板匹配相结合的配准方法，采用互信息为相似度准则进行图像的模板匹配，得到候选的匹配子图，通过待配准大图其余部分与模板和子图的空间关系，得到配准后的大图，分别计算按照各个候选子图配准情况下的待配准图的互信息，得到最大的互信息所对应的子图，确定最终的配准结果。实验结果表明，该方法在对灰度差较大的图像进行配准时，能够获得良好的效果。

关键词: 图像配准, 互信息, 模板匹配

Abstract: Aiming at the defect of local extreme of traditional mutual information method that computes both big images’ mutual information directly, a registration method that combines mutual information and template matching is presented. The template matching based on similarity criteria of mutual information is carried out, and candidates of sub-images are got, according to spatial relationship between the big image and template and spatial relationship between the big image and candidates. The panorama image after registration is got. After computing mutual information of both big images in the case of every candidate, the final registration is obtained according to maximum of mutual information of both images. Experimental results show this method is robust to images that have great gray scale difference.

Key words: image registration, mutual information, template matching

中图分类号: 

• N945