Title: RANCOR: Non-Linear Image Registration with Total Variation Regularization

Abstract: Optimization techniques have been widely used in deformable registration, allowing for the incorporation of similarity metrics with regularization mechanisms. These regularization mechanisms are designed to mitigate the effects of trivial solutions to ill-posed registration problems and to otherwise ensure the resulting deformation fields are well-behaved. This paper introduces a novel deformable registration algorithm, RANCOR, which uses iterative convexification to address deformable registration problems under total-variation regularization. Initial comparative results against four state-of-the-art registration algorithms are presented using the Internet Brain Segmentation Repository (IBSR) database.