ZT-D 2 SR: Dual-Domain Guided Adaptive Super-Resolution for remote sensing images

Abstract

Remote sensing image super-resolution aims to recover high-frequency details from low-resolution images. However, when processing complex terrains and tiny objects, existing methods often struggle with insufficient detail recovery, structural distortion, and limited generalization. Although diffusion models possess powerful generative priors, their direct application to the remote sensing domain suffers from high computational costs, false textures, and structural deformations. To address these issues, this paper proposes a Dual-Domain Guided Adaptive Super-Resolution method (D 2 SR). Unlike conventional methods that mainly optimize in a single domain, our method builds an efficient adaptive generation framework, improving robustness to complex unknown degradations while significantly improving sampling efficiency. Specifically, to mitigate geometric deformations common in diffusion models, we explicitly extract the spatial structural feature map of the low-resolution image as strong geometric guidance. This stabilizes the generation trajectory, enhances the recovery precision of sharp boundaries, and suppresses hallucinations. Furthermore, we propose a fine-grained refinement strategy based on multi-domain joint optimization. By integrating a frequency-domain loss with spatial and perceptual losses into a multidimensional composite constraint, the network is forced to precisely focus on recovering easily smoothed highfrequency details, thereby substantially boosting semantic fidelity. Extensive experiments on multiple benchmark datasets demonstrate that D 2 SR achieves an optimal balance between precise geometric reconstruction and generative detail enhancement. It not only comprehensively outperforms state-of-the-art methods in reconstruction quality but also provides a highly practical solution for large-scale remote sensing image processing with its superior inference efficiency.