基于迭代重加权的非刚性图像配准  

Iteratively Reweighted Method Based Nonrigid Image Registration

作　　者：韩雨[1] 王卫卫[1] 冯象初[1] 

机构地区：[1]西安电子科技大学理学院应用数学系,西安710071

出　　处：《自动化学报》2011年第9期1059-1066,共8页Acta Automatica Sinica

基　　金：国家自然科学基金(60872138)资助~~

摘　　要：非刚性图像配准问题是当今重要的研究课题.本文提出一类基于能量最小化方法的非刚性图像配准模型,其中包括单模态和多模态两个模型.在单模态模型中,正则项采用迭代重加权的L2范数度量,一方面克服了迭代收敛不同步的问题,另一方面使新模型既能保持图像的边缘几何结构,又能避免块效应的产生.在多模态模型中,不同模态的图像被转化为同一模态进行处理,提高了配准的效率.在模型求解方面,利用算子分裂和交替最小化的方法,将原问题转化为阈值和加性算子分裂的迭代格式进行求解.数值实验表明,本文的方法对含噪以及变形较大的图像都能实现较好的配准.Nowadays, the nonrigid image registration problem has been an important research topic. This paper proposes an energy-based framework of the nonrigid image registration, including both a one-modality model and a multi-modality model. In the one-modality model, the iteratively reweighted L2 norm is used to measure the regularization term, which brings out two advantages. Firstly, it avoids the imbalance problem of the converging speed in different regions. Secondly, it preserves the important geometric structures of an image while restrains the staircase effect. In the multi-modality model, images obtained from different modalities are converted into the one-modality ones, and then methods, handling the one-modality problems can be used to deal with the multi-modality problems. By exploiting the techniques of the operator splitting and the alternative minimization, we solve our model by shrinking and additional operator splitting （AOS）. Numerical results demonstrate that the proposed method performs well even for noisy images and images with large deformation.

关 键 词：图像配准 光流场 多模态 互信息 算子分裂 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]