基于Fourier-Mellin变换的医学图像配准算法  

Medical Image Registration Algorithm Based on Fourier-Mellin Transformation

作　　者：雷超阳 刘军华 李桥[2] 

机构地区：[1]湖南邮电职业技术学院互联网工程系,湖南长沙410015 [2]湖南涉外经济学院信息科学与工程学院,湖南长沙410215

出　　处：《吉首大学学报（自然科学版）》2017年第1期28-35,共8页Journal of Jishou University(Natural Sciences Edition)

基　　金：湖南省教育厅科学研究项目(15C1017)

摘　　要：针对基于互信息配准广泛应用于医学图像配准,存在计算量非常大、耗时长,受初始旋转和平移参数影响较大而使目标函数陷入局部极小等问题,采用Fourier-Mellin变换,先计算出2幅图像功率谱的反Fourier变换得到所对应的峰值位置,从而求得它们的相对平移,对于图像幅度谱用对数-极(Log-Polar)变换求出相对旋转角度,从而获得2幅图像配准参数,最后对图像进行配准.RTMT虽然能快速配准,但是配准精度有待提高.将RTMT和RMI结合在一起进行配准,即将RTMT获得的旋转角和平移量作为RMI的初始值,再应用RMI配准.这种方法不但计算量少而且运算简单,得出的结果精度高并且配准速度快,对于单模态图像配准和多模态配准都适用,还解决了容易陷入局部最优的问题.The mutual information-based registration （RMI） technology is widely used in medical image registration, but it requires heavy calculation and is time consuming, and since it is greatly influenced by initial rotation and translation parameters, the objective function has the problem of local minimum. In this paper,medical image registration using Fourier-Mellin transformation （RTMT） is proposed. The relative translation values between the reference and floating images are obtained by computing the in- verse Fourier transformation of the power spectrum between them to find the corresponding peak value position. In basically similar method, the relative rotation angle between them is acquired by imposing Log-polar transformation on the amplitude spectrums of the images, thus the registration parameters have been collected and finally the registering process between the two medical images are performed. Al- though RTMT can accomplish fast image registration,the registration accuracy needs to be further boos- ted. In this paper, therefore, the combination of RTMT with RMI is used to align the medical images. Namely, the translation and rotation values created by RTMT are taken as the initial registration parame- ters of RMI to deal with the alignment. This method is simple and has less amount of computation, and it can achieve precise and fast registration. It can be applied in single modality image registration and multi modality registration as well. It also solves the problem of local optimum.

关 键 词：FOURIER-MELLIN变换 医学图像 图像配准 互信息 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]