基于特征的自适应正则化配准算法*

Andriy Myronenko提出了一种自适应正则化的方法并将其应用于非刚性图像的配准，该方法在配准速度和配准精确度方面都取得了比较好的效果。但该方法对变形场初始值比较敏感，选择不当则会陷入局部极小值而不能得到理想的配准结果。为了使原始算法得到更广泛的应用，本文引入了基于特征点的粗配准方法，得到了与真实变形场更加接近的初始变形场，从而摆脱了局部极小值的困扰，得到了正确的配准结果。实验证明，改进后的算法在应用范围和配准精度上都有了提高。

Adaptive regularization registration algorithm based on feature

Andriy Myronenko proposed an adaptive regularization method and had applied it to non-rigid image registration. This method can achieve a good effect at registration speed and accuracy. But it makes a failure registration because of trapping into local minima when the initial deformable filed is chosen irrelevantly. In this paper, to using the algorithm extensively ,we gained the initial value of the deformable field by rough-registration based on feature point. Because the initial deformable field is close to the real value, the local minima can be overcome and format a correct result. By this way, the practicality and the accuracy of the algorithm have been enhanced.