The algorithm used for global registration is based on a block matching approach enabling robust registration (outliers rejection). The local registration implementation uses a cubic B-Spline parametrisation (Free-Form Deformation). All registration algorithms are based on symmetric approaches where forward and backward transformations can be optimised concurrently.
NiftyReg has been implemented for both CPU and GPU (through the use of CUDA).
Supported Data Format:
works well with: