TFSR: Terrain Factor Super Resolution

The increasing demand for high-resolution Digital Elevation Models (DEMs) across diverse fields necessitates efficient acquisition methods. High-resolution DEMs offer enhanced accuracy in terrain depiction, which is crucial for deriving terrain parameters like slope, aspect, and the hydrological index. Nevertheless, challenges in complex terrains and cost limitations hinder their widespread adoption. DEM super-resolution (DEM SR) emerges as a promising solution to enhance DEM accuracy while reducing acquisition costs. Existing methods involve super-resolving DEM data to derive terrain factors, yet this approach poses several challenges, including potential inaccuracies or artifacts hence affecting the reliability of derived terrain factors. Motivated by these challenges, we propose TFSR which is an end-to-end, supervised deep learning-based network that inputs low-resolution DEMs and outputs high-resolution terrain factors. Experimental results validate the efficacy of our approach, showcasing more minor errors in directly outputted high-resolution aspect maps compared to those generated from high-resolution DEMs.