计算机工程与应用 ›› 2013, Vol. 49 ›› Issue (11): 191-194.

• 图形图像处理 • 上一篇    下一篇

梯度与曲率变形力相结合的图像配准方法——大形变非刚性图像配准

贾义亭,祝  轩,蔺恩标,雷文娟   

  1. 西北大学 信息科学与技术学院,西安 710127
  • 出版日期:2013-06-01 发布日期:2013-06-14

Image registration by combination of gradient and curvature driven——large deformation non-rigid image registration

JIA Yiting, ZHU Xuan, LIN Enbiao, LEI Wenjuan   

  1. School of Information Science & Technology, Northwest University, Xi’an 710127, China
  • Online:2013-06-01 Published:2013-06-14

摘要: Active Demons算法是Demons算法的改进形式,其将形变配准视作扩散问题,利用牛顿作用力与反作用力思想,仅依靠梯度信息确定浮动图像的位移,在处理大形变配准问题时存在配准精度不高的弊病。将等照度线曲率作为一个控制形变的驱动力因素引入Active Demons扩散方程,建立了一个具有梯度与曲率双重驱动力相结合的非线性扩散模型(Active G&C model),并在Active G&C模型应用于大形变图像配准的算法实现过程中加入多分辨率策略,以提高大形变图像的配准精度。实验结果表明,这一模型较经典的Active Demons算法具有更好的配准性能。

关键词: Active Demons, 曲率驱动, 灰度偏移, 相似度, 多分辨率

Abstract: Active Demons is an improvement form of the Demons algorithm, which considers deformable image registration as diffusion and uses the idea of Newton’s reaction force. But it determines the displacement of the moving image only relying on gradient information that the registration correctness is not high. In this paper, it introduces the curvature into Active Demons diffusion equation, establishes a new image registration method by combination driven force of the gradient and curvature. And it adds multi-resolution approach to improve the large deformable image registration accuracy. Experimental results show that this algorithm has better performance to the classic Active Demons algorithm.

Key words: Active Demons, curvature driven, gray offset, similarity, multi-resolution