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.
出处
《微计算机信息》
2009年第34期125-127,共3页
Control & Automation
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
image registration
mutual information
genetic algorithm
simulated annealing algorithm
hybrid optimization algorithm
作者简介
唐瑜(1983-),男.湖南怀化人,桂林工学院电子与计算机系硕士研究生.主要从事图像处理的研究;通讯地址: 541004广西桂林桂林工学院电子与计算机系
叶汉民(1965-),男,桂林工学院电子与计算机系教授,主要从事自动化技术的教学与科研。