High-Quality Self-Supervised Deep Image Denoising

TL;DR

This paper introduces a self-supervised deep image denoising method that trains on unorganized corrupted images without clean references, achieving high quality and efficiency, suitable for various noise types and unknown noise parameters.

Contribution

It advances self-supervised denoising by improving training efficiency and image quality without needing clean data or explicit pairs, handling variable and unknown noise models.

Findings

Achieves state-of-the-art quality for Gaussian noise denoising

Performs well with Poisson and impulse noise

Handles variable and unknown noise parameters effectively

Abstract

We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a "blind spot" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data.