一种基于互信息的三级医学图像配准策略  

A Triple Level Registration Strategy of Medical Image Based on Mutual Information

作　　者：衣文文[1] 常发亮[1] 胡彦磊[1] 

机构地区：[1]山东大学控制科学与工程学院,山东济南250061

出　　处：《航天医学与医学工程》2008年第6期518-522,共5页Space Medicine & Medical Engineering

基　　金：国家自然科学基金(60775023);山东省自然科学基金(Z2005G03)

摘　　要：目的进一步解决基于互信息的配准算法在配准精度、速度和误配率之间的相互制约问题。方法提出一种新的配准参数搜索策略——三级配准,将平移、旋转参数分开搜索;在分析互信息配准的优点和局限性的基础上,将模糊理论中的欧几里德贴近度引入三级配准过程中。结果实验结果表明,本文提出的配准策略在保持基于互信息配准精度不变的情况下,迭代步数、配准时间和误配率均有明显改善。结论三级配准是一种能够有效协调配准精度、速度和误配率的配准策略。Objective To further solve the mutual restricted problems among precision, speed and disregistrated rate of registration algorithm based on mutual information （MI）. Methods A new strategy on the search of registration parameters, a triple level registration strategy, was proposed. It searched the translation parameters and the rotation parameters respectively. Based on analyzing the advantage and limitation of MI-based registration, Euclid closeness of Fuzzy theory was introduced into the process of the triple level registration. Results The computer simulation results demonstrated that the proposed registration method, on the condition of retaining the same precision, improved both the speed （step numbers and registration time） and the disregistrated rate. Conclusion The proposed triple level registration strategy is a registration strategy to be able to harmonize precision, speed and disregistrated rate effectively.

关 键 词：互信息 模糊梯度相似性 欧几里德贴近度 质心 三级配准 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]