改进的区域互信息和小波变换的图像配准

为提高图像配准的速度和精度，对基于区域互信息配准算法进行了改进，运用了两层小波分解策略的配准方法，小波分解得到的最顶层图像采用粒子群优化全局寻优算法，利用搜索的结果作为下一层Powell寻优方法的起点，另外，对待配准图像应用形态学方法去除噪音。针对不同分解层的特点，采用不同的测度方法，得到的顶层图像采用改进后的区域互信息为相似性测度，而底层采用归一化互信息测度和相位一致性的相结合的方法，不仅提高了速度，还克服了图像间明暗对比的影响。实验结果表明，提出的配准算法对图像噪声有较高的鲁棒性，可达到亚像素精度，在配准速度上也有了很大的提高。

Image registration based on improved region mutual information and wavelet transform

To improve speed and precision of image registration method, region mutual information is improved and two layer decomposition strategy of wavelet decomposition is researched in this paper. The top floor which has less data adopts particle swarm optimization optimal registration method, and uses the search result as the starting point of PoweU optimization method in the next layer. Otherwise, mathematical morphology is yielded to remove noise. Aiming to different similarity measure for dif- ferent characteristics of decomposition, the top floor adopts the improved regional mutual information. The normalized mutual information is employed in the bottom floor while developing the phase congruency to extract characteristics. It not only improves speed, but also overcomes the effect of contrast between light and shade in pictures. Experimental results demonstrate that this new algorithm can offer a more robust and a sub-pixel precision, and the registration time is reduced.