Multi-view Self-supervised Disentanglement for General Image Denoising

TL;DR

This paper introduces a self-supervised multi-view disentanglement method for image denoising that generalizes well to unseen noise types and outperforms supervised methods on real noise.

Contribution

It proposes a novel self-supervised framework that disentangles clean image features from noise using multiple corrupted views without needing clean images.

Findings

Outperforms prior self-supervised methods on synthetic and real noise

Achieves over 3 dB better performance than supervised methods on real noise

Demonstrates strong generalization to unseen noise types

Abstract

With its significant performance improvements, the deep learning paradigm has become a standard tool for modern image denoisers. While promising performance has been shown on seen noise distributions, existing approaches often suffer from generalisation to unseen noise types or general and real noise. It is understandable as the model is designed to learn paired mapping (e.g. from a noisy image to its clean version). In this paper, we instead propose to learn to disentangle the noisy image, under the intuitive assumption that different corrupted versions of the same clean image share a common latent space. A self-supervised learning framework is proposed to achieve the goal, without looking at the latent clean image. By taking two different corrupted versions of the same image as input, the proposed Multi-view Self-supervised Disentanglement (MeD) approach learns to disentangle the latent clean features from the corruptions and recover the clean image consequently. Extensive experimental analysis on both synthetic and real noise shows the superiority of the proposed method over prior self-supervised approaches, especially on unseen novel noise types. On real noise, the proposed method even outperforms its supervised counterparts by over 3 dB.