Morphing of Manifold-Valued Images inspired by Discrete Geodesics in Image Spaces

Sebastian Neumayer
Universität Kaiserslautern

Morphing of images is a fundamental challenge in image processing. The (time continuous) metamorphosis model originally proposed by Trouve, Younes and coworkers is a special form of morphing, where the space of images is endowed with a Riemannian metric incorporating the contribution due to transport and variations of image intensities along paths in the space of images. A variational time discretization of the metamorphosis energy functional and the Mosco--convergence of the time discrete to the timecontinuous metamorphosis model for square-integrable images in the Euclidean space are analyzed by Berkels, Effland and Rumpf. This talk addresses the morphing of manifold-valued images based on a time discrete geodesic paths model. We prove the existence of a minimizing sequence within the $L^_2$ space of images having values in a finite dimensional Hadamard manifold together with a minimizing sequence of admissible diffeomorphisms. We introduce a novel time continuous metamorphosis energy functional for images on Hadamard manifolds, which coincides with the original energy functional in case of Euclidean spaces, and the Mosco--convergence of the energy functionals on Hadamard manifolds is proven. We propose a space discrete model based on a finite difference approach on staggered grids, where we focus on the linearized elastic potential in the regularizing term. The numerical minimization alternates between i) the computation of a deformation sequence between given images via the parallel solution of certain registration problems for manifold-valued images, and ii) the computation of an image sequence with fixed first (template) and last (reference) frame based on a given sequence of deformations via the solution
of a system of equations arising from the corresponding Euler-Lagrange equation. Numerical examples give a proof of the concept of our ideas.

Joint work with A. Effland (U Graz), J. Persch (Zeiss AG), G. Steidl (TU Kaiserslautern), M. Rumpf (U Bonn)

Presentation (PDF File)

Back to Long Programs