基于特征点互信息的医学图像配准

Medical Image Registration Based on Mutual Information of Feature Points

针对互信息配准方法耗时较长、容易陷入极值的缺陷,提出采用ReNet的方法提取图像边缘,构建边缘特征点互信息能量函数,通过改进Powell算法对配准进行优化。通过实验,基于特征点互信息的医学图像配准方法与结构互信息相比,在配准准确率大致相同的情况下,配准所需时间大幅降低。

Aiming at the shortcomings of mutual information registration method,which is time-consuming and easy to fall into extreme values,it is proposed to use ReNet method to extract the edges of the image,the mutual information energy function of the edge feature points is constructed,and the registration by improving Powell algorithm is optimized. Through experiments,compared with the mutual information of medical images based on the mutual information of feature points,the registration time is greatly reduced when the registration accuracy is about the same.