灰度与形状同步匹配的非刚性配准研究  

Non-Rigid Registration with Simultaneously Matched Intensity and Shape

作　　者：林相波[1,2] 邱天爽[1] 阮素 Frederic Nicolier 

机构地区：[1]大连理工大学电子与信息工程学院,大连116024 [2]Departements de genie electrique et informatique industrielle, IUT de Troyes, Universite de Reims Champagne. Ardnne(10026 Troyes Cedex, France

出　　处：《北京生物医学工程》2009年第4期353-357,共5页Beijing Biomedical Engineering

基　　金：国家自然科学基金（30170259,30570475,60372081）、教育部博士点基金（20050141025）、大连理工大学青年教师培养基金（2007）资助

摘　　要：基于灰度的非刚性配准算法一般假设参考图像和浮动图像对应结构之间的灰度保持一致，然而在基于图谱的图像配准应用中，这种假设往往不符合实际。本文在给出一种可以同时校正灰度和形状差异的弹性配准算法的同时，针对该算法不能校正局部微小形变的弱点，提出采用自由项变换的方法进行校正以提高配准精度。配准实验基于20个IBSR真实脑部MRI图像，结果表明配准后图像与参考图像间的互相关系数得到明显提高。实验证明，本文提出的方法不仅能够同时校正形状差异和灰度变化，而且具有较高的配准质量。Intensity based non-rigid registration often assumes that the corresponding points are consistent with each other between reference image and floating image. However it is not true in some applications such as inter-subject registration or atlas based registration. In this paper, an elastic registration algorithm which can correct intensity and shape differences simultaneously is introduced. Based on the analysis of its inability in dealing with local small deformation, a post processing method using free-form transformation is suggested to increase the registration accuracy. Validation results on 20 normal brain MRI images from IBSR showed that the cross correlation between the registered image and the reference image increased obviously. It is indicated that this proposed method improved the registration accuracy as well as correct intensity and shape differences simultaneously.

关 键 词：图谱 弹性配准 自由项变换 核磁共振图像 

分 类 号：R318.04[医药卫生—生物医学工程]