摘要
基于互信息的图像配准方法具有自动化程度高、配准精度高等优点 ,已被广泛应用于医学图像的配准。但是 ,基于互信息的目标函数经常是不光滑的 ,存在许多局部极值 ,给问题的求解带来了很大的困难。本文讨论了互信息函数的多极值特性 ,并提出了一种粒子群优化算法 (particleswarmoptimization ,PSO)和Powell混合优化方法。经检验 ,这种方法能有效地克服互信息函数的局部极值 ,大大地提高了配准精度 ,达到亚像素级。
Image registration based on mutual information is of high automatization and high accuracy in registration. Hence, it has been widely exploited in medical image registration. Mutual information functions are often unsmooth. There are lots of local maximums, which abversely obstruct optimization algorithms. In this paper, features of mutual information functions are discussed and a hybrid algorithm combined by PSO and Powell is proposed. Experiments show that this hybrid algorithm could efficiently restrain local maximums of mutual information function.Also the registration accuracy could be improved to sub-pixel level.
出处
《北京生物医学工程》
2005年第1期8-12,55,共6页
Beijing Biomedical Engineering
基金
国家自然科学基金 (5 0 2 75 0 19)
教育部博士学科点专项科研基金项目 (2 0 0 10 44 10 0 5 )资助
