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 <a class="link-external link-https" href="https://github.com/CrispyFeSo4/AnySR" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve the efficiency and scalability problems in the Single Image Super - Resolution (SISR) task. Specifically, the paper proposes a new method named AnySR to achieve the SISR task of any - scale and any - resource. #### The specific problems are as follows: 1. **Limitations of fixed - scale**: - Most of the existing SISR methods can only handle a fixed magnification factor, which leads to the need to reconstruct a new super - resolution network every time to adapt to different resolution requirements, wasting computing resources. - In practical applications, there is a great demand for multi - scale SISR tasks, but the existing methods cannot adapt well to this demand. 2. **Inefficient use of resources**: - Although some methods can handle multi - scale tasks in one network, they still consume too much computing resources when handling small - scale tasks. For example, handling a smaller magnification factor (such as ×2) does not require a very complex network structure, while the existing methods usually use the same complexity to handle tasks of all scales. - This way of using resources is not flexible and efficient enough when computing resources are limited or changeable. 3. **Performance degradation caused by parameter sharing**: - In the process of trying to improve efficiency through parameter sharing, the mutual influence between different scales may lead to performance degradation. Especially when handling small - scale tasks, the performance may be affected due to the reduction of parameters. #### Solutions of AnySR: - **Any - resource implementation**: Select networks with different complexities according to the complexity of the task. Use a simpler network for smaller - scale tasks and a more complex network for larger - scale tasks. This can make more efficient use of computing resources. - **Feature - interweaving enhancement**: Insert scale information during the feature extraction process to ensure that each scale of the task can get enough attention, thus compensating for the performance loss caused by parameter sharing. Through these innovations, AnySR not only realizes the super - resolution task of any scale, but also can operate efficiently under different computing resource conditions, improving the flexibility and efficiency of the SISR task.

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

Rethinking Multi-Contrast MRI Super-Resolution: Rectangle-Window Cross-Attention Transformer and Arbitrary-Scale Upsampling

Single Remote Sensing Image Super-Resolution Via a Generative Adversarial Network with Stratified Dense Sampling and Chain Training

Activating More Information in Arbitrary-Scale Image Super-Resolution

Real-world single image super-resolution: A brief review

Meta-SR: A Magnification-Arbitrary Network for Super-Resolution

ASDN: A Deep Convolutional Network for Arbitrary Scale Image Super-Resolution

Image Super-Resolution With Cross-Scale Non-Local Attention and Exhaustive Self-Exemplars Mining

Toward Real-World Super Resolution With Adaptive Self-Similarity Mining

Multi-scale Single Image Super-Resolution with Remote-Sensing Application Using Transferred Wide Residual Network

A2M: an Amplification-Arbitrary Module for Remote Sensing Image Super-Resolution

ARM: Any-Time Super-Resolution Method

ICF-SRSR: Invertible scale-Conditional Function for Self-Supervised Real-world Single Image Super-Resolution

Learning Dual-Level Deformable Implicit Representation for Real-World Scale Arbitrary Super-Resolution

Enhanced Implicit Function-Based Network for Arbitrary-Scale Image Super-Resolution

A Feature Reuse Framework with Texture-adaptive Aggregation for Reference-based Super-Resolution

Single Image Super‐resolution Based on Progressive Fusion of Orientation‐aware Features

A Data Enhancement Method toward Real-World Single Image Super-Resolution

An Arbitrary Scale Super-Resolution Approach for 3D MR Images via Implicit Neural Representation

Single Image Super-Resolution with Multiscale Similarity Learning.

Exploiting Self-Supervised Constraints in Image Super-Resolution