Multi-Scale Recursive and Perception-Distortion Controllable Image Super-Resolution

TL;DR

This paper presents a multi-scale recursive image super-resolution method with perceptual control, achieving high perceptual quality and efficient processing, by modifying a multi-grid back-projection architecture and introducing a novel discriminator.

Contribution

It introduces a controllable super-resolution framework using a modified MGBP architecture and a multi-scale recursive discriminator with a new natural image statistics layer.

Findings

Achieved top perceptual quality rankings in PIRM Challenge 2018

Proposed a discriminator with multi-scale, recursive, and statistical features

Model uses only 281k parameters and processes images in 0.2 seconds on average

Abstract

We describe our solution for the PIRM Super-Resolution Challenge 2018 where we achieved the 2nd best perceptual quality for average RMSE<=16, 5th best for RMSE<=12.5, and 7th best for RMSE<=11.5. We modify a recently proposed Multi-Grid Back-Projection (MGBP) architecture to work as a generative system with an input parameter that can control the amount of artificial details in the output. We propose a discriminator for adversarial training with the following novel properties: it is multi-scale that resembles a progressive-GAN; it is recursive that balances the architecture of the generator; and it includes a new layer to capture significant statistics of natural images. Finally, we propose a training strategy that avoids conflicts between reconstruction and perceptual losses. Our configuration uses only 281k parameters and upscales each image of the competition in 0.2s in average.