DADSR: Degradation-Aware Diffusion Super-Resolution Model for Object-Level SAR Image

Synthetic aperture radar (SAR) plays a vital role in remote sensing applications but suffers from coupled degradation problems, including resolution deterioration, speckle noise, and defocusing effects, primarily caused by system limitations, imaging mechanisms, and nonideal target/platform motions. These intertwined degradations severely compromise image quality and subsequent interpretation reliability. Existing methods often address single-type degradation with predefined severity levels, proving inadequate for real-world scenarios involving complex coupled degradations. To overcome this limitation, we propose DADSR, a degradation-aware diffusion super-resolution model for object-level SAR images. First, a self-supervised learning framework based on contrastive learning is developed to extract discriminative degradation representations. Subsequently, a degradation-aware diffusion model is designed by incorporating the learned degradation priors, achieving adaptive super-resolution reconstruction under coupled degradation conditions. The super-resolution model based on diffusion models effectively mitigates the oversmoothing issue during the super-resolution process, while the degradation priors prevent the model from generating unnatural artifacts. Furthermore, we establish MSTAR-CD, an object-level SAR coupled degradation benchmark dataset that simulates realistic coupled degradations in the complex domain under homogeneous degradation assumptions. Experiment results demonstrate that our method attains comparable or exceeding performance to state-of-the-art models in quantitative metrics and downstream tasks.