Skip to content
Back to Projects
Space GAN Academic Research

Cloud Removal in Satellite Imagery

GAN-Based Reconstruction for Cloud-Obscured Earth Observation Data

GAN-based approach to cloud removal in satellite imagery, developed through the International Space Master Programme at the University of Luxembourg.

Type
Academic research
Institution
University of Luxembourg
Period
2019 to 2021

Research Area

Cloud cover is a persistent obstacle in Earth observation. Optical satellite imagery, used for land monitoring, agriculture, disaster response, and environmental analysis, is frequently obscured, limiting the usability of individual acquisitions. Computational approaches that can reconstruct surface information beneath cloud cover are an active area of remote sensing research.

This research, conducted through the International Space Master Programme at the University of Luxembourg (2019–2021), investigated the use of Generative Adversarial Networks (GANs) for cloud removal in satellite imagery. GANs offer a generative modelling approach to inpainting obscured regions using learned representations of surface patterns from cloud-free acquisitions.

Interested in Geospatial AI?

Get in touch to discuss how deep learning and generative models can address remote sensing and Earth observation challenges.

Contact R&D Team