JPEG Information Regularized Deep Image Prior for Denoising

TL;DR

This paper introduces a novel JPEG file size-based regularization method for deep image prior denoising, enabling effective early stopping without ground-truth images by monitoring image compression size.

Contribution

It proposes using JPEG file size as a proxy metric for noise level, improving early stopping in DIP-based denoising without requiring clean images.

Findings

JPEG file size correlates with noise levels in denoised images.

Using JPEG size as a stopping criterion improves denoising performance.

The method effectively prevents overfitting in DIP denoising.

Abstract

Image denoising is a representative image restoration task in computer vision. Recent progress of image denoising from only noisy images has attracted much attention. Deep image prior (DIP) demonstrated successful image denoising from only a noisy image by inductive bias of convolutional neural network architectures without any pre-training. The major challenge of DIP based image denoising is that DIP would completely recover the original noisy image unless applying early stopping. For early stopping without a ground-truth clean image, we propose to monitor JPEG file size of the recovered image during optimization as a proxy metric of noise levels in the recovered image. Our experiments show that the compressed image file size works as an effective metric for early stopping.