Decay2Distill: Leveraging spatial perturbation and regularization for self-supervised image denoising

TL;DR

This paper introduces Decay2Distill, a self-supervised image denoising approach that uses spatial perturbation and regularization, avoiding reliance on noise assumptions and achieving robust, cross-domain performance.

Contribution

It presents a novel unpaired self-supervised denoising method grounded in structural degradation and regularization, improving robustness over existing noise-dependent techniques.

Findings

Significant improvement over previous denoising methods

Consistent performance across different data domains

Effective without relying on noise property assumptions

Abstract

Unpaired image denoising has achieved promising development over the last few years. Regardless of the performance, methods tend to heavily rely on underlying noise properties or any assumption which is not always practical. Alternatively, if we can ground the problem from a structural perspective rather than noise statistics, we can achieve a more robust solution. with such motivation, we propose a self-supervised denoising scheme that is unpaired and relies on spatial degradation followed by a regularized refinement. Our method shows considerable improvement over previous methods and exhibited consistent performance over different data domains.