Interpretation-Oriented Cloud Removal via Observation-Anchored Residual Flow with Geo-Contextual Alignment

Researchers from the Chinese University of Hong Kong, Shenzhen proposed a new framework called Geo-Anchored Cloud Removal (GACR) for cloud removal in optical remote sensing. GACR combines physically grounded residual inversion with semantic manifold constraints from vision foundation models. The framework produces superior reconstruction quality and improves downstream task accuracy.

1sec.ai