Super-Resolution of Satellite Lidars for Forest Studies Using Diffusion Generative Models

This paper presents a novel algorithm for satellite lidar super-resolution using conditional Denoising Diffusion Probabilistic Models (DDPM) in the hyperheight data cube framework. The method employs a conditional reverse diffusion process to overcome challenges in long-range lidar systems, yielding high-resolution outputs from low-resolution inputs. Enhancing lidar products like canopy height models, it improves lidar-based modeling and facilitates advanced data analysis in fields such as environmental science, urban planning, and disaster management. The approach significantly enhances Earth's surface representation, creating research and application opportunities across disciplines. The framework's efficacy was validated in the Florida Everglades National Park, showcasing an enhancement from a 3m x 6m grid with 10m footprints to a 3m x 3m grid with 3m footprints, along with an increase in vertical resolution from 0.5m to 0.25m.