Remote Sensing Image Super-Resolution

Remote sensing image super-resolution estimates finer spatial detail from coarse satellite or aerial imagery, supporting mapping, environmental monitoring, and object detection. Methods range from CNNs and transformers to GANs, diffusion and flow models, often using multispectral bands, temporal imagery, sensor physics, or self-supervised learning when high-resolution ground truth is scarce. Pansharpening, hyperspectral and SAR reconstruction, and task-specific downscaling address different sensors and applications. A central challenge is balancing visual detail with physical and radiometric fidelity: generated textures can be plausible but inaccurate, synthetic training may not transfer to real sensors, and standard image metrics do not always predict downstream task performance.