基于PDE的自适应各向异性图像配准方法研究

针对传统光流场配准模型会造成图像模糊和细节丢失的问题，提出了一种基于偏微分方程的自适应各向异性配准模型。新模型将具有自适应性的扩散滤波方法引入图像配准，定义具有图像结构保持能力的各向异性扩散函数作为模型的正则项；数据项采用作用于亮度常量假设的非二次惩罚函数以增加模型的稳健性。实验结果表明，新模型能够有效保持图像特征，实现对大脑等复杂图像的有效配准。

Research on PDE-based adaptive anisotropic image registration

A PDE-based adaptive anisotropic model for image registration was proposed to solve the problem that traditional optical flow brings on the image blurring and details losing. The new model introduces adaptivity diffusion filter to image registration, and defines anisotropic diffusion function with the ability of preserving image structure as the regularization term; the data term use a non-quadratic penalty function with the assumption in brightness constant to improve the robust of the model. The experimental results show that the model can efficiently protect the image structure and achieve accurate registration of the complex image like brain.