Neural Universal Discrete Denoiser

TL;DR

This paper introduces Neural DUDE, a deep neural network-based universal discrete denoiser that does not require clean training data, using pseudo-labels and a novel training objective to outperform previous methods.

Contribution

The paper proposes a new framework for training neural denoisers without clean data, utilizing pseudo-labels and a specialized objective function.

Findings

Neural DUDE outperforms previous state-of-the-art denoisers.

The method effectively trains DNNs without supervised clean data.

A systematic rule for hyperparameter selection enhances practical usability.

Abstract

We present a new framework of applying deep neural networks (DNN) to devise a universal discrete denoiser. Unlike other approaches that utilize supervised learning for denoising, we do not require any additional training data. In such setting, while the ground-truth label, i.e., the clean data, is not available, we devise "pseudo-labels" and a novel objective function such that DNN can be trained in a same way as supervised learning to become a discrete denoiser. We experimentally show that our resulting algorithm, dubbed as Neural DUDE, significantly outperforms the previous state-of-the-art in several applications with a systematic rule of choosing the hyperparameter, which is an attractive feature in practice.