Stabilizing RED using the Koopman Operator

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chavan, Shraddha, Chaudhury, Kunal N.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911141393858560
author Chavan, Shraddha
Chaudhury, Kunal N.
author_facet Chavan, Shraddha
Chaudhury, Kunal N.
contents The widely used RED (Regularization-by-Denoising) framework uses pretrained denoisers as implicit regularizers for model-based reconstruction. Although RED generally yields high-fidelity reconstructions, the use of black-box denoisers can sometimes lead to instability. In this letter, we propose a data-driven mechanism to stabilize RED using the Koopman operator, a classical tool for analyzing dynamical systems. Specifically, we use the operator to capture the local dynamics of RED in a low-dimensional feature space, and its spectral radius is used to detect instability and formulate an adaptive step-size rule that is model-agnostic, has modest overhead, and requires no retraining. We test this with several pretrained denoisers to demonstrate the effectiveness of the proposed Koopman stabilization.
format Preprint
id arxiv_https___arxiv_org_abs_2509_05736
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Stabilizing RED using the Koopman Operator
Chavan, Shraddha
Chaudhury, Kunal N.
Image and Video Processing
The widely used RED (Regularization-by-Denoising) framework uses pretrained denoisers as implicit regularizers for model-based reconstruction. Although RED generally yields high-fidelity reconstructions, the use of black-box denoisers can sometimes lead to instability. In this letter, we propose a data-driven mechanism to stabilize RED using the Koopman operator, a classical tool for analyzing dynamical systems. Specifically, we use the operator to capture the local dynamics of RED in a low-dimensional feature space, and its spectral radius is used to detect instability and formulate an adaptive step-size rule that is model-agnostic, has modest overhead, and requires no retraining. We test this with several pretrained denoisers to demonstrate the effectiveness of the proposed Koopman stabilization.
title Stabilizing RED using the Koopman Operator
topic Image and Video Processing
url https://arxiv.org/abs/2509.05736