一种改进的最大互信息医学图像配准算法

给出了一种基于最大互信息和边缘互方差的医学图像配准算法. 这种算法是在计算原始图像互信息之后， 引入参考图像和浮动图像的边缘互方差值，从而建立起一个新的测度函数来指导寻优过程，最终实现图像配准. 与传统的最大互信息配准算法相比，避免了单纯依赖图像联合直方图所造成的不稳定性，使配准能更加符合图像的特点. 实验证明，这种改进算法有以下优点：配准参数曲线峰值尖锐，易于选择出最优值；在图像灰度缺失的情况下配准有较高的准确性；在噪声方差增大的情况下配准有较强的鲁棒性.

Medical registration approach using improved maximization of mutual information

Gives a new algorithm of medical image registration based on maximum of mutual information(MMI) and edge correlative deviation.This algorithm combines the edge correlative deviation of the reference image and floating image with the mutual information values of original images,and builds a new function to guide the searching course and thereby to realize the registration.Comparing to the standard MMI,this new algorithm accords with the individuality of images,by avoiding the instability brought by the mutual histogram but considering the information of the voxels' intensity values.Tests have proven that this new method additionally has several improved attributes in aspects of the more obvious climaxes of parameters' curves,more diminished errors on extremes conditions that images lack intensity values,better robust when resisting aggravated white noise.