Higher-order spatial accuracy in diffeomorphic image registration (Q889436)

Image registration is an important technique in digital image analysis in which differing images of the same object (e.g. MRI scans) are fit into a shared coordinate system so that they can be compared. Given two images, the goal is to find a suitable transformation \(T\) which deforms one image into another image. Using cost functionals to quantify the difference between the reference image and the transformed image turns image registration into an optimization problem on the space of allowed transformations. The paper under review discretizes a cost functional for image registration problems. It is shown that minima of the discretized functional can be computed in a satisfying way, and that these discretized solutions converge to solutions of the optimization problem of the original functional in dependence of the discretization parameter. The authors employ a so-called particle method, in which particles are used as a finite-dimensional representation of a diffeomorphism \(T\). They show that equipping such particles with so-called jet-data allows for achieving registrations with higher order of accuracy than ordinary particle methods are able to achieve.