摘要
研究医学图像的应用问题,针对传统的基于特征点驱动的医学图像配准需要专业医师首先找出配对的特征点,这种巨大的工作量往往会导致一些错误。在医疗中为达到配准精确要求,提出了一种全自动的特征点驱动的医学颅脑CT图像配准方法,采用SIFT和形状因子来产生配对的特征点,并构成特征点的局部和区域特征向量,后用薄板样条插值(TPS)方法实施两幅图像的配准。仿真结果表明,在保证配准精度的基础上显著提高了算法的速度。
There are two categories of image registration algorithms: one is based on matching corresponding landmarks and the other is based on matching intensity information.The traditional landmarks-based image registration must ask professional medical image doctor to found corresponding landmarks for registration,and its heavy workload always leads to some errors.This former is used in registration of brain CT slices in this paper.Firstly,the SIFT algorithm and shape factor are used to produce landmarks.Then,a primary data set which includes both the local and global feature vectors is constructed by using these landmarks.Finally,the registration is carried out on the data set by thin-plate spline(TPS) interpolation.Simulation showed that the proposed method accelerates the registration procedure obviously.
出处
《计算机仿真》
CSCD
北大核心
2011年第5期279-282,共4页
Computer Simulation
基金
国家自然科学基金(60771007)
关键词
医学图像配准
形状因子
薄板样条插值
Medical image registration
Shape factor
Thin-plate spline interpolation