Super-Resolution for Cross-Sensor Optical Remote Sensing Images

Generative adversarial network (GAN) models are becoming popular in the field of remote sensing for generating high spatial resolution images from their low resolution versions. In this study, four models including two basic Super-resolution GAN models and two non-GAN Deep Learning models were trained and tested to achieve 2.5m, and 5m spatial resolution from their 10m spatial resolution satellite data. The comparison of results showed that the SRGAN model performed better than the other deep learning models. The performance metrics were also found to be consistent with available literature.