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

@article{Cohen2021RegularizationBD,
  title={Regularization by Denoising via Fixed-Point Projection (RED-PRO)},
  author={Regev Cohen and Michael Elad and Peyman Milanfar},
  journal={SIAM J. Imaging Sci.},
  year={2021},
  volume={14},
  pages={1374-1406}
}
Inverse problems in image processing are typically cast as optimization tasks, consisting of data fidelity and stabilizing regularization terms. A recent regularization strategy of great interest utilizes the power of denoising engines. Two such methods are the Plug-and-Play Prior (PnP) and Regularization by Denoising (RED). While both have shown state-of-the-art results in various recovery tasks, their theoretical justification is incomplete. In this paper, we aim to enrich the understanding… 
An Interpretation Of Regularization By Denoising And Its Application With The Back-Projected Fidelity Term
TLDR
This paper shows that the RED gradient can be seen as a (sub)gradient of a prior function--but taken at a denoised version of the point, and proposes to combine RED with the Back-Projection (BP) fidelity term rather than the common Least Squares (LS) term that is used in previous works.
Fixed-Point and Objective Convergence of Plug-and-Play Algorithms
TLDR
A novelty in this regard is that the analysis covers non-symmetric denoisers such as nonlocal means and almost any convex data-fidelity, and makes use of the convergence theory of averaged operators and a special inner product derived from the linear denoiser.
Learning Maximally Monotone Operators for Image Recovery
TLDR
An operator regularization is performed, where a maximally monotone operator (MMO) is learned in a supervised manner, and a universal approximation theorem proving that nonexpansive NNs provide suitable models for the resolvent of a wide class of MMOs is provided.
Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors
TLDR
This work proposes a new asynchronous RED (ASYNC-RED) algorithm that enables asynchronous parallel processing of data, making it significantly faster than its serial counterparts for large-scale inverse problems.
Single Image Non-uniform Blur Kernel Estimation via Adaptive Basis Decomposition
TLDR
This paper proposes a general, non-parametric model for dense non-uniform motion blur estimation and shows that it contributes to bridging the gap between model-based and data-driven approaches for deblurring real photographs.
Feasibility-based fixed point networks
TLDR
This work fuses data-driven regularization and convex feasibility in a theoretically sound manner using feasibility-based fixed point networks (F-FPNs), which demonstrate performance increases when compared to standard TV-based recovery methods for CT reconstruction and a comparable neural network based on algorithm unrolling.
CoIL: Coordinate-based Internal Learning for Imaging Inverse Problems
We propose Coordinate-based Internal Learning (CoIL) as a new deep-learning (DL) methodology for continuous representation of measurements. Unlike traditional DL methods that learn a mapping from the
Gradient Step Denoiser for convergent Plug-and-Play
TLDR
This work proposes a new type of PnP method, based on half-quadratic splitting, for which the denoiser is realized as a gradient descent step on a functional parameterized by a deep neural network.
ASYNC-RED: A PROVABLY CONVERGENT ASYN-
TLDR
This work proposes a new asynchronous RED (ASYNC-RED) algorithm that enables asynchronous parallel processing of data, making it significantly faster than its serial counterparts for large-scale inverse problems.
Accelerating Plug-and-Play Image Reconstruction via Multi-Stage Sketched Gradients
TLDR
A novel multi-stage sketched gradient iterations which first perform downsampling dimensionality reduction in the image space, and then efficiently approximate the true gradient using the sketches gradient in the low-dimensional space is proposed.
...
1
2
3
...

References

SHOWING 1-10 OF 117 REFERENCES
Image Restoration by Iterative Denoising and Backward Projections
  • Tom Tirer, R. Giryes
  • Computer Science, Mathematics
    IEEE Transactions on Image Processing
  • 2019
TLDR
An alternative method for solving inverse problems using off-the-shelf denoisers, which requires less parameter tuning and is competitiveness with task-specific techniques and the P&P approach for image inpainting and deblurring is proposed.
Regularization by Denoising: Clarifications and New Interpretations
TLDR
This work proposes a new framework called Score-Matching by Denoising (SMD), which aims to match a “score” (i.e., the gradient of a log-prior) and shows tight connections between SMD, kernel density estimation, and constrained minimum mean-squared error denoising.
Learning Proximal Operators: Using Denoising Networks for Regularizing Inverse Imaging Problems
TLDR
This paper studies the possibility of replacing the proximal operator of the regularization used in many convex energy minimization algorithms by a denoising neural network, and obtains state-of-the-art reconstruction results.
Block Coordinate Regularization by Denoising
TLDR
A new block coordinate RED algorithm is developed that decomposes a large-scale estimation problem into a sequence of updates over a small subset of the unknown variables and theoretically analyzes the convergence of the algorithm and discusses its relationship to the traditional proximal optimization.
The Little Engine That Could: Regularization by Denoising (RED)
TLDR
This paper provides an alternative, more powerful, and more flexible framework for achieving Regularization by Denoising (RED): using the denoising engine in defining the regulariza...
Performance Analysis of Plug-and-Play ADMM: A Graph Signal Processing Perspective
  • Stanley H. Chan
  • Computer Science
    IEEE Transactions on Computational Imaging
  • 2019
TLDR
This paper restricts the denoisers to the class of graph filters under a linearity assumption, or more specifically the symmetric smoothing filters, and introduces a new analysis technique via the concept of consensus equilibrium to provide interpretations to problems involving multiple priors.
Plug-and-Play ADMM for Image Restoration: Fixed-Point Convergence and Applications
TLDR
It is shown that for any denoising algorithm satisfying an asymptotic criteria, called bounded denoisers, Plug-and-Play ADMM converges to a fixed point under a continuation scheme.
A Convergent Image Fusion Algorithm Using Scene-Adapted Gaussian-Mixture-Based Denoising
TLDR
This work proposes using a scene-adapted denoiser plugged into the iterations of the alternating direction method of multipliers, which yields state-of-the-art results but also allows proving convergence of the resulting algorithm.
DeepRED: Deep Image Prior Powered by RED
TLDR
This work proposes to bring-in the concept of Regularization by Denoising (RED), which leverages existing denoisers for regularizing inverse problems, and shows how the two can be merged into a highly effective unsupervised recovery process while avoiding the need to differentiate the chosen denoiser, and leading to very effective results.
Boosting the Performance of Plug-and-Play Priors via Denoiser Scaling
TLDR
A scaling parameter is introduced that adjusts the magnitude of the denoiser input and output so that the performance of plug-and-play prirs for denoising CNN priors that do not have explicitly tunable parameters is improved.
...
1
2
3
4
5
...