Super-Resolution of Sentinel-2 Images Based on Deep Channel-Attention Residual Network
Xi Zhu, Yang Xu, Zhihui Wei · 2019
Sentinel-2 data has become an important tool for current and future earth observation due to its high quality, free availability and world-wide coverage. However, some of the spectral bands are sensed at reduced resolution due to design considerations and sensor hardware limitations. So in this paper we present a super-resolution method based on Convolutional Neural Networks (CNNs) to infer all the 20m spectral bands in the highest available resolution. This is accomplished by using an improved residual network and meanwhile we propose a channel attention mechanism to adaptively rescale the characteristics of the channels by considering the interdependencies among the channels. The proposed solution compares against several alternative methods according to different quality indexes. Our network provides the best results and a compelling visual effect on the sentinel-2 images.