Lightweight Arbitrary-Scale Super-Resolution of Remote Sensing Images via Super-Scale Feature

Remote sensing image (RSI) super-resolution (SR) demands lightweight and efficient methods due to required rapid response in practical applications. Integrating RSIs with different resolutions for diverse applications also requires arbitrary-scale SR, making fix-scaled SR scale inflexible. Therefore, a lightweight SR algorithm capable of arbitrary-scale is necessary for RSIs. To address the above issue, a super-scale feature-based lightweight arbitrary-scale (SFLA) SR network is proposed in this paper. The network consists of two modules: 1) A super-scale feature extraction (SSFE) module that extracts features at both the initial low-resolution (LR) and an integer super-scale resolution, 2) A self-attention implicit function reconstruction (SIFR) module that utilizes multi-layer perceptron (MLP) network and self-attention mechanism for pixel-wise feature mapping to achieve superior SR results. Comparative experiments and ablation results demonstrate that the proposed SFLA algorithm effectively strikes a good balance between performance and complexity.

I. Introduction

The spatial resolution of remote sensing image (RSI) significantly influences image quality and the ability to identify ground targets. Obtaining higher resolution RSIs requires more complex equipment and advanced technical specifications, leading to higher costs. In contrast, image super-resolution (SR) offers a practical advantage of obtaining high-resolution (HR) RSIs with lower costs under similar hardware conditions. Furthermore, for the sake of long-term temporal change analysis, image SR can make "outdated" low-resolution (LR) RSIs renewal again and be used together with those latest high-resolution RSIs. Therefore, SR holds significant importance for RSIs processing.