基于多分辨率和互信息的图像配准方法研究

用互信息进行图像配准可使配准的精度达到亚像素级,但该方法因计算量很大致使配准速度较慢.为了提高图像配准速度,并进一步提高配准的精度,首先运用多分辨率的方法将两幅待配准图像分别分解成n个不同分辨率的子图像,然后利用互信息先将分辨率最低的图像进行粗配准,并将此配准的结果作为下一个分辨率较高的图像配准的基础,继续这个过程直至最高分辨率图像被配准,即运用由粗到精的配准策略减小计算量,从而提高配准速度.用Visual C++6.0编程实现该配准方法,实验结果表明在图像配准过程中将多分辨率和互信息结合起来使用,配准速度得到很大的提高,并且由于多分辨率的使用有利于避免局部最大值,使得配准精度也有了一定程度的提高.

Method Research of Image Registration Based on Multi-Resolution and Mutual Information

The accuracy of registration can reach as high a level as sub-pixel if the mutual information method was applied to image registration.But the speed of the registration method is slow for there are too much data to calculate.In order to improve the speed and accuracy to a certain extent,the multi-resolution approach and the coarse-to-fine strategy was used this paper when we perform image registration.About all,we get n images which are on different scales from the two images that they will be matched.Then the lowest resolution image will be matched at first,the next registration will be base on the last time match result.This process is repeated until the highest resolution image has been matched.The registration method was programmed using visual c++6.0 by the author.Result of the experiment indicates that the registration speed is improved in a great degree for cutting down the calculation,and the accuracy is improved to a certain degree for the multi-resolution approach is benefit to avoid the local maximum.