基于互信息相似性度量的多时相遥感图像配准

图像配准是多源图像分析的基础，特别是对于多时相遥感图像的变化检测，图像配准的性能直接影响到变化检测的实现。针对当前利用互信息实现图像配准时出现的稳健性问题，本文提出一种新的多时相遥感图像配准方法，在计算互信息时引入图像的梯度信息，作为多时相遥感图像配准的相似性度量准则，并在配准过程中应用多分辨率分析技术，使得图像配准在提高稳健性的同时减少运算量。将此方法用于变化检测中的多时相遥感图像配准实验，结果显示该方法比直接应用灰度互信息的配准方法具有更好的稳健性。

Image Registration Based on Mutual Information for Multi-Temporal Remote Sensing

Image registration approaches based on mutual information have been applied for many areas such as medical image analysis and remote sensing. However, the robust of registration based on mutual information is a problem, especially for multi-temperal remote sensing images where there exist changes between multi-temperal. In order to improve the robust of image registration, a new registration approach based on mutual information was proposed in this paper. At first, the gradient field was established using wavelet transformation, then, it was combined into the mutual information to compute the similar measure of image registration. The transform parameters between multi-temporal images were attained according to maximum of mutual information. In order to accelerate the registration, coarse to fine multi-resolution search strategy had been used. Apply our registration approach to multi-temperal remote sensing images, the results show that our registration approach can perform fairly well.