Towards Optimized Mars Terrain Reconstruction Methods for Virtual Reality

Virtual Reality enables the exploration of realistic outer space simulations, where accurate extraterrestrial terrain models are essential features. Due to extreme data retrieval challenges, such models often contain missing values that can hinder the 3D scenarios. We present a surface reconstruction method based on generative diffusion models, and compare it with other void-filling techniques in the task of restoring degraded Martian terrains. Reconstruction outputs were quantitatively assessed by comparing their error and structural consistency with the original values. Our method achieved the best performance on most metrics, suggesting promising prospects for extraterrestrial terrain visualization.