Regularization of Inverse Problems via Time Discrete Geodesics in Image Spaces

TL;DR

This paper introduces a novel regularization method for inverse imaging problems that combines a discrete geodesic path model with an $L^2$-$TV$ variational approach, ensuring stable solutions and demonstrating effectiveness in tomography and superresolution.

Contribution

It develops a new regularization framework combining geodesic paths and $L^2$-$TV$ models, with proven stability and novel minimization algorithms.

Findings

Stable solutions depend continuously on data.

Effective in sparse and limited angle tomography.

Successful in superresolution tasks.

Abstract

This paper addresses the solution of inverse problems in imaging given an additional reference image. We combine a modification of the discrete geodesic path model for image metamorphosis with a variational model,actually the $L^2$-$TV$ model, for image reconstruction. We prove that the space continuous model has a minimizer which depends in a stable way from the input data. Two minimization procedures which alternate over the involved sequences of deformations and images in different ways are proposed. The updates with respect to the image sequence exploit recent algorithms from convex analysis to minimize the $L^2$-$TV$ functional. For the numerical computation we apply a finite difference approach on staggered grids together with a multilevel strategy. We present proof-of-the-concept numerical results for sparse and limited angle computerized tomography as well as for superresolution demonstrating the power of the method.