Abstract: The medical image registration based on mutual information is a high-precision registration method，which can reach sub-pixel accuracy with no need to extract anatomical features from images.But it can not meet the need of real-time clinic due to its low speed.This paper proposes a speed program：a rough registration to narrow the range of mutual information search，a method of gray-scale compression based on a gray histogram for speeding up mutual information computation，a hybrid optimization algorithm of genetic algorithm and simplex method for convergence acceleration.Experiments show that the modified method can significantly reduce the registration time.

Key words: image registration, mutual information, rough registration, gray-scale compression, genetic algorithm, simplex method

摘要： 基于互信息的医学图像配准是一种高精稳健的自动配准算法，可以达到亚像素级精度且无需提取解剖特征而倍受重视，但其最大问题是速度慢，致使其不能满足临床的实时需求。在分析影响其速度因素的基础上提出一套加速方案，即采用快速粗配准来缩小互信息的搜索范围、利用非等间隔的灰度压缩来加快互信息的计算、通过混合遗传算法和单纯形算法来加快互信息的搜索。实验表明，改进后的算法在保证配准精度的前提下能显著提高配准速度。

关键词: 图像配准, 互信息, 粗配准, 灰度压缩, 遗传算法, 单纯形法