摘要
基于图像几何特征的配准方法速度快,应用价值高,但由于几何特征提取的不准确问题使配准易陷入局部极小,而目前常用的全局优化算法又存在收敛速度慢的问题。该文提出了一种以图像边界的平均Haudorff距离作为代价函数,基于多分辨率分析和混合优化策略的图像配准方法,将其用于医学图像配准,并与基于Hausdorff距离的几种传统方法进行分析比较,实验结果显示,在模拟退火算法和Powell算法的混合优化策略下,新方法具有良好的全局优化性能和时间性能。
Image registration method based on geometric features is fast and available in application to medical image registration. At the same time, due to inaccuracy of geometric features extracted from images, local optimum is often obtained. Moreover, global optimization algorithms are slower. This paper presents a new image registration method based on mean Hausdofff distance, multi-resolution, and hybrid optimization to be applied to medical image registration, and it is compared with other methods. Results of experiments show that the new method is excellent both in global optimization and rantime.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2006年第15期194-196,共3页
Computer Engineering
基金
国家自然科学基金资助项目(60573019)
广东省自然科学基金资助项目(05103541
31340)
