A Comprehensive Survey of Super-Resolution Remote Sensing Image Datasets: Evolution, Challenges, and Future Directions

High-resolution (HR) imagery plays a crucial role in remote sensing applications such as land-cover classification, environmental monitoring, urban planning, and disaster assessment, yet acquiring such imagery remains challenging due to sensor limitations, atmospheric disturbances, and high data costs. Super-resolution (SR) has emerged as a promising computational alternative, enhancing spatial detail from low-resolution data without requiring hardware upgrades. Despite significant progress in SR algorithms—particularly those driven by deep learning—the supporting datasets have received comparatively little attention. This survey presents a comprehensive review of over 50 publicly available SR datasets, including satellite multispectral, panchromatic, hyperspectral, UAV, and airborne imagery. We systematically analyze them by image quantity, spatial resolution, sensor diversity, acquisition platform, and land-cover representation. We categorize these datasets based on SR application type (single-image vs. multi-image, pan-involve image, multi-spectral image, and hyperspectral image) and explore how dataset characteristics influence algorithm development and evaluation. Several challenges are identified, including the prevalence of unrealistic degradation models, geometric misalignments, land-class imbalances, and trade-offs between resolution and geographic coverage. In response, we propose a roadmap for next-generation SR datasets that prioritizes multi-source, multi-angle, and multi-temporal observations, standardized preprocessing pipelines, and balanced representation of land types. This work aims to inform dataset selection, support fair benchmarking, and guide the design of robust and scalable SR models for real-world remote sensing scenarios.