Self-adaptive Anisotropy Registration Model in Image Registration Based on Partial Differential Equation

Jianjun WU

Hunan City University, Yiyang, Hunan, 413000, China

Abstract: As the optical flow motion estimation algorithm in image registration appears serious image blur, large errors, error messages and other issues, so this paper presents a PDE-based self-adaption anisotropy registration

model. The model can adaptively choose the diffusion rate of tangent direction and normal direction according to the local image structure information, to protect the edge of the image in the registration process not to be blurred, at the same time, designs strong data item and improve the robustness of the algorithm. Experimental results show that: this model can make the image clear and the details are not easy to lose, and achieve effective registration.

Keywords: Image Feature; Evolution; Penalty Function; Gradient Modulus