A Hybrid Scheme for Efficient Deformable Image Registration.

Deformable image registration (DIR) is one of the important processing steps for image analysis in the compute vision field. Although various methods have been proposed to deal with the DIR problem, it is still challenging to achieve a good balance between the registration speed and accuracy. This paper introduces a novel hybrid registration model which is capable of fast and accurately warping a target image to a reference image. This achievement owns to a coarse-to-fine registration scheme, where an automatic thin plate splines-based (A-Tps) method is proposed for preliminary registration, then an improved inertia Demons (I-Demons) algorithm is developed for fine warping the result of previous output. Experimental results demonstrate the outstanding performance of the proposed model in terms of registration speed and accuracy.