Multi-scale frequency separation network for image deblurring

Image deblurring aims to restore the detailed texture information or structures from blurry images, which has become an indispensable step in many computer vision tasks. Although various methods have been proposed deal with image problem, most of them treated as a whole and neglected characteristics different frequencies. In this paper, we present new method called multi-scale frequency separation network (MSFS-Net) for deblurring. MSFS-Net introduces module (FSM) into encoder-decoder architecture capture low- high-frequency at multiple scales. Then, simple cycle-consistency strategy sophisticated contrastive learning (CLM) are respectively designed retain low-frequency recover during At last, features scales fused by cross-scale feature fusion (CSFFM). Extensive experiments on benchmark datasets show that achieves state-of-the-art performance.