Permanent link to Research Commons versionhttps://hdl.handle.net/10289/14738
Super-resolution of satellite imagery poses unique challenges. We propose a hybrid method comprising two existing deep network super-resolution approaches, namely a feedforward network called DeepSUM, and ESRGAN, a GAN-based approach, to super-resolve multiple low-resolution images by a factor of four to obtain a single high-resolution image. We also introduce a novel loss function, called variation loss, to better define edges and textures to create a sharper, perceptually better output. Using our hybrid, we inherit some of the advantages of both deep learning approaches, resulting in super-resolved images that better show boundaries, textures, and details.
This is an author’s accepted version of an article published in the Proceedings of 2021 36th International Conference on Image and Vision Computing New Zealand (IVCNZ). ©2021 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.