Latent topic-based super-resolution for remote sensing

Abstract

This letter presents a novel single-image Super-Resolution (SR) approach based on latent topics specially designed to remote sensing imagery. The proposed approach pursues to super-resolve topics uncovered from low-resolution images instead of super-resolving image patches themselves. An experimental comparison is con- ducted using nine di ff erent SR methods over four aerial image data- sets. Experiments revealed the potential of latent topics in remote sensing SR by reporting that the proposed approach is able to provide a competitive advantage especially in low noise conditions.