Abstract: Demons algorithm considers image registration as diffusion and gained displacement of the moving image only by gradient. Because the deformation directions of image are uncertain when the gradient is very small, the registration results will be wrong. In order to solve this problem, this paper considers the curvature can be introduced into Demons diffusion equation to establish a new image registration method by combination of the gradient and curvature driven. Experiments show that this algorithm is an effective image registration algorithm.

Key words: Demons, curvature, gray offset, similarity

摘要： 针对Demons算法将形变配准视作扩散问题，仅靠梯度信息确定浮动图像的位移，在梯度非常小时图像形变方向不能确定，会导致错误的配准变换这一弊病，将等照度线曲率作为一个控制形变的驱动力因素引入Demons扩散方程，建立了一个具有梯度与曲率双重驱动力的非线性扩散模型，将其应用于图像配准。实验结果表明，这一模型较经典Demons算法具有更好的配准性能。

关键词: Demons, 曲率驱动, 灰度偏移, 相似度