Non-Rigid Registration Based on Huber Prior: Application to Multiphase Liver CT Volumes Alignment

This study investigates a new non-rigid registration method based on Huber prior. The goal of image registration is to determine a reasonable transformation, such that a transformed version of the first image is similar to the second one. In order to improve the computing efficiency and registration precision, we model the motion of the object by both affine transformation and B-spline based Free-Form deformation. Our innovation is using Huber prior as penalty term of the energy function for better results. The FFD model with Huber prior which has high accuracy and strong robustness can settle the over-smoothing and edge information deficiency problem. We applied the proposed algorithm to contrast-enhanced liver CT volumes registration. Distinctly, the experiment results show that our method gets better results compared to affine transformation and original FFD.