摘要
提出一种基于拉普拉斯谱的医学图像配准算法,通过将谱图理论应用到医学图像配准中并引入特征向量,以达到提高配准精确度和计算效率的目的。该算法根据医学图像的解剖特征来构造拉普拉斯矩阵,通过分析拉普拉斯矩阵的谱得到匹配关系;采用射影变换模型,计算射影矩阵;通过坐标变换和图像插值方法实现图像配准。实验结果表明,该算法与经典的最大互信息配准算法相比,提高了单传感器和多传感器医学图像配准的精度,并且降低运算复杂度。
The medical image registration algorithm based on Laplacian spectrum algorithm is proposed. The idea of using graph theory as medical image point registration is proposed. By eigenvector, the higher registration precision and the strong ability have been achieved. The Laplacian spectrum of the anatomy feature point sets of the two images is constructed to match. The strategy uses the homogeneous matrix, coordinate transfer and the image interpolation to register. Experimental results show that it has more efficiency and higher registration precision than maximization of mutual information whenever in the single sensor or multi-sensor.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第14期231-232,235,共3页
Computer Engineering
基金
国家自然科学基金资助项目(60772121)
安徽省教育厅重点科研计划基金资助项目(KJ2010A021)
安徽大学"211工程"学术创新团队基金资助项目
关键词
配准
拉普拉斯谱
图
互信息
传感器
registration
Laplacian spectrum
graph
mutual information
sensor
