LSwinSR: UAV Imagery Super-Resolution Based on Linear Swin Transformer

Super-resolution, which aims to reconstruct high-resolution (HR) images from low-resolution (LR) images, has drawn considerable attention and has been intensively studied in computer vision and remote sensing communities. Super-resolution technology is especially beneficial for unmanned aerial vehicles (UAVs), as the number and resolution of images captured by UAVs are highly limited by physical constraints such as flight altitude and load capacity. In the wake of the successful application of deep learning methods in the super-resolution task, in recent years, a series of super-resolution algorithms have been developed. In this article, for the super-resolution of UAV images, a novel network based on the state-of-the-art Swin Transformer is proposed with better efficiency and competitive accuracy. Meanwhile, as one of the essential applications of the UAV is land cover and land use monitoring, simple image quality assessments such as the peak-signal-to-noise ratio (PSNR) and the structural similarity index measure (SSIM) are not enough to comprehensively measure the performance of an algorithm. Therefore, we further investigate the effectiveness of super-resolution methods using the accuracy of semantic segmentation. The code is available at https://github.com/lironui/GeoSR.

I. Introduction

With the continuous development in both hardware reliability and control strategy, UAVs have been widely employed in more and more practical applications [1], such as atmosphere monitoring [2], tracking and surveillance [3], and land cover and land use monitoring [4]. However, limited by the physical characteristics of the UAV such as the flight altitude and load capacity, sometimes, it is impractical to continuously obtain HR images through UAV, especially for those applications that need large-scale and long-duration UAV video data [5]. Super-resolution is a promising solution to alleviate this dilemma, which can reconstruct HR images from LR observations [6]. In fact, the super-resolution technology particularly the deep-learning-based methods has already been widely applied for the processing of various kinds of images, such as natural images [7], [8], medical images [9], [10], and remote sensing images [11], [12], [13]. These works can not only enhance the image quality but also can further facilitate downstream applications [14], [15].