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