Regularization by Denoising: Clarifications and New Interpretations

  title={Regularization by Denoising: Clarifications and New Interpretations},
  author={Edward T. Reehorst and Philip Schniter},
  journal={IEEE Transactions on Computational Imaging},
Regularization by denoising (RED), as recently proposed by Romano, Elad, and Milanfar, is powerful image-recovery framework that aims to minimize an explicit regularization objective constructed from a plug-in image-denoising function. [] Key Method To explain the RED algorithms, we propose a new framework called Score-Matching by Denoising (SMD), which aims to match a “score” (i.e., the gradient of a log-prior).

Figures and Tables from this paper

An Interpretation Of Regularization By Denoising And Its Application With The Back-Projected Fidelity Term
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.
Regularization by Denoising via Fixed-Point Projection (RED-PRO)
This paper reformulates RED as a convex optimization problem utilizing a projection (RED- PRO) onto the fixed-point set of demicontractive denoisers, and establishes a novel unification of RED-PRO and PnP, while providing guarantees for their convergence to the globally optimal solution.
On the Construction of Averaged Deep Denoisers for Image Regularization
This work proposes to construct averaged (contractive) image denoisers by unfolding ISTA and ADMM iterations applied to wavelet denoising and demonstrates that their regularization capacity for PnP and RED can be matched with CNN denoiser.
Learned Gradient of a Regularizer for Plug-and-Play Gradient Descent
It is shown that it is possible to train a denoiser along with a network that corresponds to the gradient of its regularizer, and this gradient of the regularizer is used in gradient-based optimization methods and obtained better results comparing to other generic Plug-and-Play approaches.
Monotonically Convergent Regularization by Denoising
Simulations on image deblurring and compressive sensing recovery from random matrices show the stability of MRED even when the traditional RED algorithm diverges, whose convergence does not require nonexpansiveness of the deep denoising prior.
Proximal denoiser for convergent plug-and-play optimization with nonconvex regularization
It is shown that this gradient denoiser can actually correspond to the proximal operator of another scalar function and be exploited in the nonconvex setting to obtain convergence results for PnP-PGD (Proximal Gradient Descent) and PNP-ADMM (Alternating Direction Method of Multipliers).
Block Coordinate Regularization by Denoising
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.
Solving RED via Weighted Proximal Methods
A general framework named weighted proximal methods (WPMs) is applied to address RED efficiently and it is shown that one can obtain a more efficient variant of WPMs for handling RED if an effective weighting is set.
Boosting the Performance of Plug-and-Play Priors via Denoiser Scaling
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.
Online Deep Equilibrium Learning for Regularization by Denoising
This work theoretically analyze ODER giving insights into its convergence and ability to approximate the traditional DEQ approach, and numerical results suggest the potential improvements in training/testing complexity due to ODER on three distinct imaging applications.


The Little Engine That Could: Regularization by Denoising (RED)
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...
The SURE-LET Approach to Image Denoising
It is shown that a denoising algorithm merely amounts to solving a linear system of equations which is obviously fast and efficient, and the very competitive results obtained by performing a simple threshold on the undecimated Haar wavelet coefficients show that the SURE-LET principle has a huge potential.
Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising
This paper investigates the construction of feed-forward denoising convolutional neural networks (DnCNNs) to embrace the progress in very deep architecture, learning algorithm, and regularization method into image Denoising, and uses residual learning and batch normalization to speed up the training process as well as boost theDenoising performance.
Scene-Adapted Plug-and-Play Algorithm with Guaranteed Convergence: Applications to Data Fusion in Imaging
This paper proposes a PnP approach where a scene-adapted prior is plugged into ADMM (alternating direction method of multipliers), and proves convergence of the resulting algorithm.
Plug-and-Play priors for model based reconstruction
This paper demonstrates with some simple examples how Plug-and-Play priors can be used to mix and match a wide variety of existing denoising models with a tomographic forward model, thus greatly expanding the range of possible problem solutions.
Local Kernels That Approximate Bayesian Regularization and Proximal Operators
This paper broadly connects kernel-based filtering with general variational formulations of Bayesian regularized least squares and the related concept of proximal operators to establish how one can approximate the solution of the resulting global optimization problems using locally adaptive filters with specific kernels.
A Tour of Modern Image Filtering: New Insights and Methods, Both Practical and Theoretical
  • P. Milanfar
  • Computer Science
    IEEE Signal Processing Magazine
  • 2013
A practical and accessible framework is presented to understand some of the basic underpinnings of algorithms in wide use such as block-matching and three-dimensional filtering (BM3D), and methods for their iterative improvement (or nonexistence thereof) are discussed.
Majorization–Minimization Algorithms for Wavelet-Based Image Restoration
Several new results are proved which strongly support the claim that the SI does not compromise the usefulness of this class of algorithms, and introduce a new algorithm, resulting from using bounds for nonconvex regularizers, which confirms the superior performance of this method, when compared to the one based on quadratic majorization.
Image Restoration using Autoencoding Priors
This work builds on the key observation that the output of an optimal denoising autoencoder is a local mean of the true data density, and uses the magnitude of this mean shift vector as the negative log likelihood of the natural image prior for image restoration.
Acceleration of RED via Vector Extrapolation