Deformable Groupwise Image Registration using Low-Rank and Sparse Decomposition
- A new groupwise distance measure for non-parametric image registration based on low-rank and sparse decompositions, inspired by Heber and Pock, 2014
- Regularization through Total Variation or Curvature Penalty
- Optimization is performed using iterative relinearization and the primal-dual hybrid gradient algorithm from Chambolle and Pock, 2011
Add your local copy of this repository to your MATLAB path. Then, simply run
Demo/demo_mf_nn_tv_registration_no_ref_ml.m
for a demo on synthetic image data with TV regularization.
As a method of comparison, a simple variance-based groupwise registration method as in Metz et al., 2011 is implemented as well. A demo is found in
Demo/demo_var_tv_registration_no_ref_ml.m