基于局部互信息和自动更新的医学图像形变不均匀配准

摘要

针对互信息只考虑图像像素的灰度信息和图像存在形变不均匀的情况下,本文提出局部互信息和部分多层次B样条结合的方法.第一步采用主轴质心法对多模医学图像进行粗配准,从而快速实现两幅图像的粗配准.第二步采用部分多层次B样条法针对解决局部形变不均匀的配准对象,首先是粗网格进行全局粗配准,然后只是对部分区域实现细化网格处理,加快配准速度.文中对网格进行自动更新,将采用以局部互信息为相似度检测,结合这3种方法,从而实现多模医学图像的精确和快速配准.%Based on that the mutual information only considers the gray information of image pixels and the heterogeneous deformation of images,authors introduce the method combining local mutual information with partial multilevel B spline.The principal axes and centroid based method is used for the coarse registration of multi-mode medical images,so as to quickly achieve the coarse registration of two images.And then,partial multilevel B spline method is used to deal with the registration object of local heterogeneous deformation (performing global coarse registration for coarse grid,and achieving grid refinement on parts of the region,and accelerating the speed of registration).Herein,the mesh is updated automatically,and the local mutual information is used as the similarity detection.Combining these three methods,we can realize the accurate and fast registration of multi-mode medical images.