Reference-Based Multi-Level Features Fusion Deblurring Network for Optical Remote Sensing Images

Blind image deblurring is a long-standing challenge in remote sensing restoration tasks. It aims to recover latent sharp from blurry while the blur kernel unknown. To solve this problem, many priors-based algorithms and learning-based have been proposed. However, most of these methods are based on single image. Due lack high frequency information, images restored by still some deficiencies edge texture details. In work, we propose novel deep learning model named Reference-Based Multi-Level Features Fusion Deblurring Network (Ref-MFFDN), which registers reference multi-level feature space transfers high-quality textures registered features assist deblurring. Comparative experiments testing set prove that our Ref-MFFDN outperforms state-of-the-art approaches both quantitative evaluation visual results, indicates effectiveness using More ablation demonstrates robustness input size, fusion network (MFFN) effect different levels multi-feature extractor (MFE) algorithm performance.