Learning local regularization for variational image restoration

TL;DR

This paper introduces a novel framework that learns a local regularization function for image restoration tasks using a neural network trained with Wasserstein GANs, applicable to denoising and deblurring.

Contribution

It presents a new method to learn local regularizers directly from data, adaptable to various image restoration problems, using a critic-based approach with unpaired data.

Findings

Effective in denoising and deblurring tasks

Outperforms traditional regularization methods

Flexible and applicable to different restoration problems

Abstract

In this work, we propose a framework to learn a local regularization model for solving general image restoration problems. This regularizer is defined with a fully convolutional neural network that sees the image through a receptive field corresponding to small image patches. The regularizer is then learned as a critic between unpaired distributions of clean and degraded patches using a Wasserstein generative adversarial networks based energy. This yields a regularization function that can be incorporated in any image restoration problem. The efficiency of the framework is finally shown on denoising and deblurring applications.