Stable Backward Diffusion Models that Minimise Convex Energies

@article{Bergerhoff2020StableBD,
  title={Stable Backward Diffusion Models that Minimise Convex Energies},
  author={Leif Bergerhoff and Marcelo C{\'a}rdenas and J. Weickert and M. Welk},
  journal={Journal of Mathematical Imaging and Vision},
  year={2020},
  volume={62},
  pages={941 - 960}
}
The inverse problem of backward diffusion is known to be ill-posed and highly unstable. Backward diffusion processes appear naturally in image enhancement and deblurring applications. It is therefore greatly desirable to establish a backward diffusion model which implements a smart stabilisation approach that can be used in combination with an easy-to-handle numerical scheme. So far, existing stabilisation strategies in the literature require sophisticated numerics to solve the underlying… Expand
Networks for Nonlinear Diffusion Problems in Imaging
TLDR
This work develops a network architecture based on nonlinear diffusion processes, named DiffNet, that is well suited for diffusion-related problems in imaging and obtains competitive results to the established U-Net architecture. Expand

References

SHOWING 1-10 OF 71 REFERENCES
Modelling Stable Backward Diffusion and Repulsive Swarms with Convex Energies and Range Constraints
TLDR
This work derives a class of one-dimensional backward evolutions from the minimisation of strictly convex energies, and stabilises these inverse evolutions by imposing range constraints, to establish a comprehensive theory for the time-continuous evolution, and to prove a stability condition for an explicit time discretisation. Expand
Theoretical Foundations for Discrete Forward-and-Backward Diffusion Filtering
TLDR
It is shown that for a novel nonstandard space discretisation which pays specific attention to image extrema, one can apply a modification of the space-discrete well-posedness and scale-space framework of Weickert (1998) which allows to establish well-posingness and a maximum-minimum principle for the resulting dynamical system. Expand
A Discrete Theory and Efficient Algorithms for Forward-and-Backward Diffusion Filtering
TLDR
A comprehensive theory for space-discrete and time-continuous FAB diffusion processes is established and it is shown that this theory carries over to the fully discrete case, when an explicit time discretisation with a fairly restrictive step-size limit is applied. Expand
An Efficient and Stable Two-Pixel Scheme for 2D Forward-and-Backward Diffusion
TLDR
This work extends the explicit nonstandard scheme by Welk et al. (2009) from the 1D scenario to the practically relevant two-dimensional setting and proves that under a fairly severe time step restriction, this 2D scheme preserves a maximum–minimum principle. Expand
Stable explicit time marching in well-posed or ill-posed nonlinear parabolic equations
This paper analyses an effective technique for stabilizing pure explicit time differencing in the numerical computation of multidimensional nonlinear parabolic equations. The method uses easilyExpand
Image segmentation and edge enhancement with stabilized inverse diffusion equations
TLDR
A family of first-order multidimensional ordinary differential equations with discontinuous right-hand sides with stabilized inverse diffusion equations (SIDEs) is introduced and existence and uniqueness of solutions, as well as stability, are proven for SIDEs. Expand
Stabilized Richardson leapfrog scheme in explicit stepwise computation of forward or backward nonlinear parabolic equations
Abstract The numerical computation of ill-posed, time-reversed, nonlinear parabolic equations presents considerable difficulties. Conventional stepwise marching schemes for such problems, whetherExpand
Two stable methods with numerical experiments for solving the backward heat equation
This paper presents results of some numerical experiments on the backward heat equation. Two quasi-reversibility techniques, explicit filtering and structural perturbation, to regularize theExpand
Compensating operators and stable backward in time marching in nonlinear parabolic equations
Linear and nonlinear parabolic advection–dispersion–reaction equations play a significant role in the geophysical sciences, where they are widely used in modeling contaminant transport in subsurfaceExpand
Solution of inverse diffusion problems by operator-splitting methods
In this paper the inverse solution of the general (non-linear) diffusion problem or backward heat conduction problem. It is assumed that the direct solution can be satisfactorily modelled, forExpand
...
1
2
3
4
5
...