MGBPv2: Scaling Up Multi-Grid Back-Projection Networks

TL;DR

This paper presents MGBPv2, an advanced multi-grid back-projection network for super-resolution that is scalable, adaptable, and capable of processing very large images, achieving top perceptual quality in a challenge.

Contribution

MGBPv2 introduces scalable, parameter-flexible modifications to the multi-grid back-projection network, enabling high-quality super-resolution for large images and improved generality over its predecessor.

Findings

Won 1st in perceptual quality in AIM-2019 challenge

Achieved 5th place in high-fidelity (PSNR)

Capable of processing 8K images with overlapping patches

Abstract

Here, we describe our solution for the AIM-2019 Extreme Super-Resolution Challenge, where we won the 1st place in terms of perceptual quality (MOS) similar to the ground truth and achieved the 5th place in terms of high-fidelity (PSNR). To tackle this challenge, we introduce the second generation of MultiGrid BackProjection networks (MGBPv2) whose major modifications make the system scalable and more general than its predecessor. It combines the scalability of the multigrid algorithm and the performance of iterative backprojections. In its original form, MGBP is limited to a small number of parameters due to a strongly recursive structure. In MGBPv2, we make full use of the multigrid recursion from the beginning of the network; we allow different parameters in every module of the network; we simplify the main modules; and finally, we allow adjustments of the number of network features based on the scale of operation. For inference tasks, we introduce an overlapping patch approach to further allow processing of very large images (e.g. 8K). Our training strategies make use of a multiscale loss, combining distortion and/or perception losses on the output as well as downscaled output images. The final system can balance between high quality and high performance.