Accelerating FAIR on GPUs

Non-rigid image registration tries to find a suitable deformation field between two images that map one image into the other. This is widely used in medical imaging applications. Here, it is important to have fast image registration algorithms, e.g. in order to reduce the waiting time for the patient or to be able to register images during a surgery. We show how compute intensive parts of the image registration algorithm can be parallelized and ported to GPUs and the issues pertaining how the GPU code can be coupled to the FAIR (Flexible Algorithms for Image Registration) software package.