Best-Buddy GANs for Highly Detailed Image Super-Resolution

TL;DR

This paper introduces Best-Buddy GANs (Beby-GAN), a novel approach for highly detailed single image super-resolution that allows flexible patch supervision and region-aware learning, improving detail generation especially in textured areas.

Contribution

The paper proposes Beby-GAN, which relaxes the one-to-one mapping constraint and employs region-aware adversarial training for enhanced detail synthesis in super-resolution.

Findings

Beby-GAN outperforms existing methods in producing detailed textures.

The region-aware strategy improves focus on textured regions.

Constructed a 4K ultra-high-resolution dataset for super-resolution research.

Abstract

We consider the single image super-resolution (SISR) problem, where a high-resolution (HR) image is generated based on a low-resolution (LR) input. Recently, generative adversarial networks (GANs) become popular to hallucinate details. Most methods along this line rely on a predefined single-LR-single-HR mapping, which is not flexible enough for the SISR task. Also, GAN-generated fake details may often undermine the realism of the whole image. We address these issues by proposing best-buddy GANs (Beby-GAN) for rich-detail SISR. Relaxing the immutable one-to-one constraint, we allow the estimated patches to dynamically seek the best supervision during training, which is beneficial to producing more reasonable details. Besides, we propose a region-aware adversarial learning strategy that directs our model to focus on generating details for textured areas adaptively. Extensive experiments justify the effectiveness of our method. An ultra-high-resolution 4K dataset is also constructed to facilitate future super-resolution research.