Single-Image Super-Resolution for Remote Sensing Data Using Deep Residual-Learning Neural Network

Abstract

Single image super-resolution (SISR) plays an important role in remote sensing image processing. In recent years, deep convolutional neural networks have achieved state-of-the-art performance in the SISR field of common camera images. Although the SISR method based on deep learning is effective on general camera images, it is not necessarily effective on remote sensing images because of the significant difference between remote sensing images and common camera images. In this paper, the VDSR network (proposed by Kim et al. in 2016) was found to be invalid for Sentinel-2A remote sensing images; we then proposed our own neural network, which is called the remote sensing deep residual-learning (RS-DRL) network. Our network achieved better performance than VDSR on Sentinel-2A remote sensing images.