摘要
针对传统的均匀子采样的最小生成树配准方法对采样率敏感,导致配准鲁棒性降低的问题,提出了一种融合梯度信息的最小生成树医学图像配准算法.该算法首先提取均匀子采样点集,并在此基础上构造最小生成树;然后使用最小生成树来估计Rényi熵;最后将图像间的边缘梯度信息融入到配准框架中.通过在公共数据集RREP上,与传统的基于均匀子采样的最小生成树配准算法和基于归一化互信息配准算法相比,提出的算法在达到良好配准精度的同时,具有更平滑的配准函数和较强的鲁棒性.
The conventional MST（minimum spanning tree）image registration based on uniform sub-sampling is so sensitive to sampling rate that the registration robustness is thus reduced.To solve the problem,MST integrated with gradient information is proposed as a medical image registration algorithm.In the algorithm,the uniform sub-sampling point set is extracted to form MST,which is used to estimate the Rényi entropy directly.As a result,the edge gradient information between images is integrated into the registration framework.Comparison results of the images obtained from Vanderbilt retrospective registration project（RREP）showed that the algorithm proposed can provide smoother registration function and better robustness than both the MST registration algorithm based on conventional uniform sub-sampling and the registration algorithm based on normalized mutual information（NMI）.
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2010年第10期1393-1396,共4页
Journal of Northeastern University(Natural Science)
基金
辽宁省重大科技计划项目(2008402001)
关键词
医学图像配准
最小生成树
Rényi熵
图像梯度
medical image registration
minimum spanning tree（MST）
Rényi entropy
image gradient
