Abstract: Medical image registration has important clinical implications and has been a hotspot in medical image processing field.Image registration methods based on mutual information criteria have been widely used in three-dimensional medical image registration for its automation and high accuracy.However，the problems of large volume of data and long processing time appear at the same time.In this paper，normalized mutual information measure and a new orthogonal array optimization technique are applied to 3-D registration of MR image sequence of human brain to shorten the processing time.Compared with traditional optimization algorithm，the experimental results show that the proposed method can significantly improve the registration rate while remain the registration accuracy.

Key words: medical image registration, mutual information, normalized mutual information, damping orthogonal array, multi-
resolution strategy

摘要： 医学图像配准具有重要的临床应用价值，并一直是医学图像处理领域的热点研究问题。基于互信息的配准方法由于自动化程度高和配准精度高的优点而被广泛应用于三维医学图像配准;但是，也存在着数据量大、计算速度慢的问题。采用归一化互信息测度，并将一种新的正交优化技术应用于颅脑MR影像序列的三维配准，旨在缩短处理时间。将该方法与使用传统优化算法的配准方法作了比较，实验结果表明，提出的方法能够显著提高配准速度，且不降低配准精度。

关键词: 医学图像配准, 互信息, 归一化互信息, 阻尼正交表, 多分辨率策略