Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real Image Super-Resolution

TL;DR

This paper introduces SRResCycGAN, a cyclic GAN-based deep learning model for real-world image super-resolution that maintains domain consistency and performs well on real images, surpassing traditional bicubic methods.

Contribution

The paper proposes a novel cyclic GAN framework for real image super-resolution, addressing domain mismatch issues in traditional methods.

Findings

Achieves comparable results to state-of-the-art methods on AIM 2020 dataset.

Generalizes well to real-world images beyond clean training data.

Easy deployment on mobile and embedded devices.

Abstract

Recent deep learning based single image super-resolution (SISR) methods mostly train their models in a clean data domain where the low-resolution (LR) and the high-resolution (HR) images come from noise-free settings (same domain) due to the bicubic down-sampling assumption. However, such degradation process is not available in real-world settings. We consider a deep cyclic network structure to maintain the domain consistency between the LR and HR data distributions, which is inspired by the recent success of CycleGAN in the image-to-image translation applications. We propose the Super-Resolution Residual Cyclic Generative Adversarial Network (SRResCycGAN) by training with a generative adversarial network (GAN) framework for the LR to HR domain translation in an end-to-end manner. We demonstrate our proposed approach in the quantitative and qualitative experiments that generalize well to the real image super-resolution and it is easy to deploy for the mobile/embedded devices. In addition, our SR results on the AIM 2020 Real Image SR Challenge datasets demonstrate that the proposed SR approach achieves comparable results as the other state-of-art methods.