Abstract:Regularization is necessary when solving inverse problems to ensure the well-posedness of the solution map. Additionally, it is desired that the chosen regularization strategy is convergent in the sense that the solution map converges to a solution of the noise-free operator equation. This provides an important guarantee that stable solutions can be computed for all noise levels and that solutions satisfy the operator equation in the limit of vanishing noise. In recent years, reconstructions in inverse problems are increasingly approached from a data-driven perspective. Despite empirical success, the majority of data-driven approaches do not provide a convergent regularization strategy. One such popular example is given by iterative plug-and-play (PnP) denoising using off-the-shelf image denoisers. These usually provide only convergence of the PnP iterates to a fixed point, under suitable regularity assumptions on the denoiser, rather than convergence of the method as a regularization technique, thatis under vanishing noise and regularization strength. This paper serves two purposes: first, we provide an overview of the classical regularization theory in inverse problems and survey a few notable recent data-driven methods that are provably convergent regularization schemes. We then continue to discuss PnP algorithms and their established convergence guarantees. Subsequently, we consider PnP algorithms with learned linear denoisers and propose a novel spectral filtering technique of the denoiser to control the strength of regularization. Further, by relating the implicit regularization of the denoiser to an explicit regularization functional, we are the first to rigorously show that PnP with a learned linear denoiser leads to a convergent regularization scheme. The theoretical analysis is corroborated by numerical experiments for the classical inverse problem of tomographic image reconstruction.

Averaged Deep Denoisers for Image Regularization

Averaged Deep Denoisers for Image Regularization

Convergent Regularization in Inverse Problems and Linear Plug-and-Play Denoisers

Monotonically Convergent Regularization by Denoising

The Little Engine that Could: Regularization by Denoising (RED)

Regularization by Denoising via Fixed-Point Projection (RED-PRO)

Instabilities in Plug-and-Play (PnP) algorithms from a learned denoiser

Learning Deep CNN Denoiser Prior for Image Restoration

On the Strong Convexity of PnP Regularization Using Linear Denoisers

Dilated Residual Encode-Decode Networks for Image Denoising

Convergent Plug-and-Play with Proximal Denoiser and Unconstrained Regularization Parameter

An Adaptive Image Denoising Method Based on Deep Rectified Denoising Auto-Encoder

Plug-and-Play Image Reconstruction Is a Convergent Regularization Method

Constructing an Interpretable Deep Denoiser by Unrolling Graph Laplacian Regularizer

Plug-and-Play image restoration with Stochastic deNOising REgularization

Learning pseudo-contractive denoisers for inverse problems

A relaxed proximal gradient descent algorithm for convergent plug-and-play with proximal denoiser