基于最大互信息的多源图像配准研究

Multi-source image registration based on maximization of mutual information

对同一场景使用不同传感器获取的多幅源图像往往存在许多差异,若要对其进行像素级图像融合,则首先必须对多幅源图像进行配准,配准精度要求达到像素级。本文在对多源图像配准理论和方法深入分析的基础上,对基于最大互信息的图像配准算法及其实现进行了深入研究,并将该方法应用于多源图像配准,通过仿真实验证明采用基于最大互信息的图像配准方法对来自不同传感器的多源图像进行配准,可以达到像素级的配准精度,该精度标准符合像素级图像融合对源图像的配准精度要求。

There are a lot of differences in multi-source images from various sensors about the same scene. When these images are fused in pixel-level, they should be registered first and registration precision is required in pixellevel. In this paper multi-source image registration theory and methods are discussed and analyzed deeply. Based on this, we make an emphasis on the image registration algorithm by Maximization of Mutual Information （MMI） and its implementation. At the same time, we use this method to register the multi-source images such as remote-sensing images and medical images. Experimental results show that the registration method by MMI is used to register multi- source images from various sensors can obtain registration precision in pixel-level and meet precision demand of pixel-level image fusion.