改进的基于最大归一化互信息的医学图像配准算法  

Improved Medical Image Registration Algorithm Based on Maximization of Normalized Mutual Information

作　　者：刘洪[1] 李良荣[1] 王蓓[1] 

机构地区：[1]贵州大学电子科学与信息技术学院,贵阳550025

出　　处：《贵州大学学报（自然科学版）》2008年第3期284-288,共5页Journal of Guizhou University:Natural Sciences

摘　　要：作者提出了一种基于最大归一化互信息的医学图像配准算法。该算法利用改进的部分体积插值法进行插值计算,有效地克服了图像配准中常见的局部极值问题。该算法利用最大互信息作为目标函数,POWELL优化搜索算法搜索一个最大互信息量从而获得最佳配准参数。实验证明,该算法计算简单,配准速度快,具有更好的精确性和鲁棒性。In this paper a medical image registration algorithm based on maximization of normalized mutual information was proposed. The algorithm uses the improved partial volume distribution interpolation , effectively overcome the local extremum problem which in common image registration. The algorithm uses the maximization mutual information as the objective function and uses the POWELL algorithm search one maximization mutual information results show that the proposed algorithm is simple to get the best registration parameters. Experiment , fast, and can achieve better accuracy and robustness.

关 键 词：归一化互信息 医学图像配准 部分体积插值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]