Preconditioned Plug-and-Play ADMM with Locally Adjustable Denoiser for Image Restoration

DOI10.1137/22M1504809MaRDI QIDQ6173529zbMATH OpenOpenAlexFDO

Authors Mikaël le Pendu, Christine Guillemot

Publication date 21 July 2023

Published in SIAM Journal on Imaging Sciences (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2110.00493

zbMATH Keywords

ADMMdenoisinginterpolationinverse problemsplug-and-play priorpreconditioning

Mathematics Subject Classification ID

Image processing (compression, reconstruction, etc.) in information and communication theory (94A08)

Abstract: Plug-and-Play optimization recently emerged as a powerful technique for solving inverse problems by plugging a denoiser into a classical optimization algorithm. The denoiser accounts for the regularization and therefore implicitly determines the prior knowledge on the data, hence replacing typical handcrafted priors. In this paper, we extend the concept of plug-and-play optimization to use denoisers that can be parameterized for non-constant noise variance. In that aim, we introduce a preconditioning of the ADMM algorithm, which mathematically justifies the use of such an adjustable denoiser. We additionally propose a procedure for training a convolutional neural network for high quality non-blind image denoising that also allows for pixel-wise control of the noise standard deviation. We show that our pixel-wise adjustable denoiser, along with a suitable preconditioning strategy, can further improve the plug-and-play ADMM approach for several applications, including image completion, interpolation, demosaicing and Poisson denoising.

Recommendations

• PnP-ReG: Learned Regularizing Gradient for Plug-and-Play Gradient Descent
• The Little Engine that Could: Regularization by Denoising (RED)
• Provably Convergent Plug-and-Play Quasi-Newton Methods
• Learning maximally monotone operators for image recovery

Cites work

• rm K-SVD: An Algorithm for Designing Overcomplete Dictionaries for Sparse Representation
• A proximal point analysis of the preconditioned alternating direction method of multipliers
• Algorithms and software for total variation image reconstruction via first-order methods
• Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising
• Distributed optimization and statistical learning via the alternating direction method of multipliers
• Image compression with anisotropic diffusion
• Iterative Joint Image Demosaicking and Denoising Using a Residual Denoising Network
• Preconditioned alternating direction method of multipliers for inverse problems with constraints
• Single-Image Super-Resolution via Linear Mapping of Interpolated Self-Examples
• THE TRANSFORMATION OF POISSON, BINOMIAL AND NEGATIVE-BINOMIAL DATA
• Waterloo Exploration Database: New Challenges for Image Quality Assessment Models

Cited in

(5)

• PnP-ReG: Learned Regularizing Gradient for Plug-and-Play Gradient Descent
• Provably Convergent Plug-and-Play Quasi-Newton Methods
• Averaged deep denoisers for image regularization
• Plug and Play splitting techniques for Poisson image restoration
• Theoretical characterization of effect of masks in snapshot compressive imaging

This page was built for publication: Preconditioned Plug-and-Play ADMM with Locally Adjustable Denoiser for Image Restoration

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6173529)