基于互信息的同模态医学图像配准  

Mono-modality image registration based on modified mutual information

作　　者：翟海亭[1] 吴晓娟[1] 李元建[1] 徐东[1] 

机构地区：[1]山东大学信息科学与工程学院,济南市250100

出　　处：《医疗卫生装备》2006年第6期23-24,27,共3页Chinese Medical Equipment Journal

摘　　要：以改进的互信息为测度,用Powell优化算法实现了同模态医学图像的配准。主要改进了互信息和归一化互信息的公式,减小了互信息的计算量;改进了Powell算法,保持了原搜索方向并有效避免了局部极值;用预设旋转量的方法解决了同模态医学图像配准的局部极值问题,同时,还讨论了Powell收敛阈值对配准速度的影响。实验表明,用改进的算法进行同模态图像配准可以达到亚像素精度,且在速度上有了明显提高,是1种鲁棒性强、速度快、精度高的方法。With modified mutual information （MI） measure, Powell search algorithm is utilized to realize mono-modality image registration. The MI＇s formulae are improved to reduce computation cost, and the replace of Powell＇s search direction is also rectified to keep the original search direction and eliminate local maxima. With rotation set in advance, local maxima can be avoided. The influence on registration velocity by Powell convergence threshold is also discussed. The results show that the modified MI measure can reach sub-pixel precision with obviously accelerated speed. It has good robustness, high speed and high precision.

关 键 词：同模态医学图像 图像配准 互信息 POWELL算法 

分 类 号：R318[医药卫生—生物医学工程]