Single Image Super-Resolution using Information Augmentation

Spatial Super-Resolution (SR) of images is applied in a multitude of domains. These include enhancing the resolution of old images, microscopic images, telescopic images for remote sensing and astronomical observations. This paper focuses on single image spatial resolution problem using deep learning technique in remotely sensed images. In particular, edge augmentation technique is used as an augmentation technique to increase information content, following which, SR task is carried out using a modification of one of the standard deep learning models – Enhanced Deep Super Resolution (EDSR). The obtained results are compared to the original model to assess the improvement in performance.