Automatic Tuning of Denoising Algorithms Parameters Without Ground Truth
Authors:
- Arthur FloquetAFArthur Floquet
- Sayantan DuttaSDSayantan Dutta
- Emmanuel SoubiesESEmmanuel Soubies
- Duong Hung PhamDPDuong Hung Pham
- Denis KouameDKDenis Kouame
Preprint
Abstract
Denoising is omnipresent in image processing. It is usually addressed with algorithms relying on a set of hyperparameters that control the quality of the recovered image. Manual tuning of those parameters can be a daunting task, which calls for the development of automatic tuning methods. Given a denoising algorithm, the best set of parameters is the one that minimizes the error between denoised and ground-truth images. Clearly, this ideal approach is unrealistic, as the ground-truth images are unknown in practice. In this work, we propose unsupervised cost functions -- i.e., that only require the noisy image -- that allow us to reach this ideal gold standard performance. Specifically, the proposed approach makes it possible to obtain an average PSNR output within less than 1% of the best achievable PSNR.Keywords
All Related Versions
- Version 1, 2024-01-18, ArXiv
- Published version: IEEE Signal Processing Letters, 31 (10709908), 381.
This publication has 12 references indexed in Scilit:
- A Novel Image Denoising Algorithm Using Concepts of Quantum Many-Body TheorySignal Processing, 2022
- Learning regularization parameters of inverse problems via deep neural networksInverse Problems, 2021
- Quantum Mechanics-Based Signal and Image Representation: Application to DenoisingIEEE Open Journal of Signal Processing, 2021
- Noisy-as-Clean: Learning Self-Supervised Denoising From Corrupted ImageIEEE Transactions on Image Processing, 2020
- An Overview of Overfitting and its SolutionsJournal of Physics: Conference Series, 2019
- Total Variation Denoising Via the Moreau EnvelopeIEEE Signal Processing Letters, 2017
- Stein Unbiased GrAdient estimator of the Risk (SUGAR) for Multiple Parameter SelectionSIAM Journal on Imaging Sciences, 2014
- Automatic Parameter Selection for Denoising Algorithms Using a No-Reference Measure of Image ContentIEEE Transactions on Image Processing, 2010
- Monte-Carlo Sure: A Black-Box Optimization of Regularization Parameters for General Denoising AlgorithmsIEEE Transactions on Image Processing, 2008
- Image Denoising by Sparse 3-D Transform-Domain Collaborative FilteringIEEE Transactions on Image Processing, 2007