Powell和SA混合优化的互信息图像配准  

Medical Image Registration of Mutual Information based on Powell and SA Hybrid Algorithm

作　　者：唐瑜[1] 叶汉民[1] 孔维娟[1] 

机构地区：[1]桂林工学院电子与计算机系,广西桂林541004

出　　处：《微计算机信息》2009年第34期125-127,共3页Control & Automation

摘　　要：基于互信息的图像配准方法具有鲁棒性强、配准精度高等优点,已被广泛应用于医学图像的配准。但计算互信息涉及大量的浮点运算,且搜索最大互信息时容易陷入局部极值。本文提出了一种基于Powell算法和模拟退火算法的混合优化方法。实验结果显示,该优化算法提高了搜索全局最优解的速度,限制了局部极值,而且提高了配准精度,达到亚像素级。Image registration based on mutual information is of strong robustness and high accuracy in registration.Hence,it has been widely exploited in medical image registration.But getting the mutual information of images needs a mass of floatpoint calculation,and it would easily get the local maximums. In this paper,a hybrid algorithm combined by Powell and SA is proposed.Experiments show that this hybrid algorithm could efficiently get the globally optimal solution and restrain local maximums of mutual information function.Also the registration accuracy could be improved to sub-pixel leve1.

关 键 词：图像配准 互信息 鲍威尔算法 模拟退火算法 混合优化算法 

分 类 号：R318[医药卫生—生物医学工程]