AnySR: Realizing Image Super-Resolution as Any-Scale, Any-Resource

TL;DR

AnySR is a novel framework that enables arbitrary-scale and resource-efficient single-image super-resolution, outperforming existing methods in efficiency while maintaining comparable performance.

Contribution

We introduce AnySR, a method that transforms arbitrary-scale super-resolution into an any-resource implementation, reducing resource use for smaller scales without extra parameters.

Findings

Achieves resource-efficient arbitrary-scale super-resolution.

Performs on par with existing arbitrary-scale SISR methods.

Validated on five popular SISR datasets.

Abstract

In an effort to improve the efficiency and scalability of single-image super-resolution (SISR) applications, we introduce AnySR, to rebuild existing arbitrary-scale SR methods into any-scale, any-resource implementation. As a contrast to off-the-shelf methods that solve SR tasks across various scales with the same computing costs, our AnySR innovates in: 1) building arbitrary-scale tasks as any-resource implementation, reducing resource requirements for smaller scales without additional parameters; 2) enhancing any-scale performance in a feature-interweaving fashion, inserting scale pairs into features at regular intervals and ensuring correct feature/scale processing. The efficacy of our AnySR is fully demonstrated by rebuilding most existing arbitrary-scale SISR methods and validating on five popular SISR test datasets. The results show that our AnySR implements SISR tasks in a computing-more-efficient fashion, and performs on par with existing arbitrary-scale SISR methods. For the first time, we realize SISR tasks as not only any-scale in literature, but also as any-resource. Code is available at https://github.com/CrispyFeSo4/AnySR.