摘要
差异进化算法是一种简单、可靠和有效的全局优化算法。本文在差异进化算法的基础上,将对等学习方法应用到种群初始化和进化中,提出一种基于对等差异进化算法的图像配准方法。该方法在种群进化过程中由对等跃变率决定是否生成对等种群,文中对不同对等跃变率及动态对等跃变率的不同情况进行三维医学图像配准实验。实验结果表明,只要选取合适的对等跃变率,该方法比传统差异进化算法的图像配准具有更高的精度和更好的稳定性;并且线性递减跃变率的对等差异进化配准算法比固定对等跃变率和线性递增跃变率的配准更精确、更稳定。
The differential evolution is a simple,reliable and efficient global optimization algorithm.This paper proposes an opposition-based differential evolution for the image registration.The opposition-based learning is used for the population initialization and the generation evolution based on the differential evolution.Different constant and time varying jumping rates,for determining whether generate corresponding opposition population or not,are tried on 3D medical image registration.Experimental results show that the algorithm for choosing a proper jumping rate outperforms the conventional differential evolution on the precision and the stability for medical image registration;and the linearly decreasing jumping rate performs better than constant settings and linearly increasing policy.
出处
《数据采集与处理》
CSCD
北大核心
2010年第3期358-363,共6页
Journal of Data Acquisition and Processing
关键词
差异进化
对等学习法
图像配准
differential evolution
opposition-based learning
image registration
