Abstract: Multi-resolution 3D medical image registration algorithms using hybrid strategy have been implemented widely.However，these hybrid algorithms only focus on the mixture of optimization algorithm.The influence of resolution on registration based on first-order Mutual Information（MI） or second-order MI is studied.Then，the improved strategy with hybrid similarity metric and hybrid optimization algorithm is introduced for two-level multi-resolution registration，which adoptes more suitable similarity metric for each level.Results demonstrate that the overall registration accuracy of the improved method outperforms that of method based on single similarity metric and achieved the subvoxel accuracy.In addition，the speed of the proposed algorithm is faster than that of second-order MI，slightly slower than first-order MI.

Key words: hybrid strategy, medical image registration, Mutual Information（MI）, second-order Mutual Information（MI）, Powell algorithm, simulated annealing

摘要： 基于混合策略的多分辨率算法是当前3D医学图像刚体配准中普遍采用的方法，不过其仅仅是优化算法的混合。通过研究不同分辨率对一阶互信息（常称为互信息）和二阶互信息配准的影响，在二级多分辨率策略的配准中，各级采用相对更适合的相似性测度，提出了混合优化算法和混合测度的改进算法。实验表明，改进算法在配准精度上达到了亚体素级，且明显优于基于单一测度的算法，在配准速度上远远快于基于二阶互信息单一测度的算法，略慢于基于一阶互信息单一测度的算法。

关键词: 混合策略, 医学图像配准, 互信息, 二阶互信息, 鲍威尔（Powell）算法, 模拟退火法