Regularization in Medical Image Registration using Global Linear Optimization
(English)Manuscript (preprint) (Other academic)
Common problems in image registration include having large parts of the images contain noisy, uncertain, missing or impossible motion. Regularization is the field that aims to overcome these problems. In this article, we propose a novel framework : Global Linear Optimization (GLO) which we demonstrate has the capabilities to simultaneously and globally regularize with respect to : (1) anisotropic certainty of prior motion field, (2) sliding of organ boundaries and (3) incompressibility of organ interiors. The power of the presented framework consists of being able to spatially adapt which subsets of the data each constraint should affect and then solve a large sparse linear equations system which automatically propagates a solution over the data set through an overlapping localized metric. We demonstrate the validity of the methods and the power of the GLO framework on relevant test cases and on medical data from the DIR-lab.
Keywords—Image Registration, Medical Image Analysis, Regularization, Adaptive Filtering, Medical Atlases, Global Methods, Optimization, Global Linear Optimization, Structure Tensor, Anisotropic Filtering, Partial Differential Equations
Medical Image Processing Other Computer and Information Science
IdentifiersURN: urn:nbn:se:liu:diva-117140OAI: oai:DiVA.org:liu-117140DiVA: diva2:805940