Wavelet-based variational deformable registration for ultrasound

The use of ultrasound as an imaging modality has many clinical applications. Robust and efficient image registration techniques are desirable in many of those applications but, because of the inherent artifacts in ultrasound, the registration problem is challenging. We propose a novel deformable registration method using a discrete wavelet transform (DWT) for ultrasound images. Energy maps were constructed from the details of the DWT in the multiresolution pyramid. The energy maps were then registered through minimizing and energy functional, using the calculus of variations, from coarser to more detailed levels of the pyramid. We applied the method to both simulated synthetic images and human liver ultrasound images. The experiments showed the advantages of the proposed method as an efficient and noise-robust registration framework.