Binary recovery via phase field regularization for first traveltime tomography
This work addresses the binary recovery problem in traveltime tomography for geophysical imaging, but the contribution is incremental as it applies existing phase field techniques to a new inverse problem.
The paper proposes a phase field regularization method for binary recovery of slowness in first traveltime tomography, achieving convergent discretization and validation through numerical experiments on various test problems.
We propose a double obstacle phase field methodology for binary recovery of the slowness function of an Eikonal equation found in first traveltime tomography. We treat the inverse problem as an optimization problem with quadratic misfit functional added to a phase field relaxation of the perimeter penalization functional. Our approach yields solutions as we account for well posedness of the forward problem by choosing regular priors. We obtain a convergent finite difference and mixed finite element based discretization and a well defined descent scheme by accounting for the non-differentiability of the forward problem. We validate the phase field technique with a $Γ$ - convergence result and numerically by conducting parameter studies for the scheme, and by applying it to a variety of test problems with different geometries, boundary conditions, and source - receiver locations.