This paper introduces a two-stage framework, the Distorted Underwater Image Restoration Network (DR-Net), to address the complex degradation of underwater images caused by turbulence, fluctuations, and optical properties of water. The first stage employs the Distortion Estimation Network (DE-Net), leveraging a fusion of Transformer and U-Net architectures, to estimate distortion information from degraded images and focuses on image distortion recovery. Subsequently, the Image Restoration Generative Adversarial Network (IR-GAN) in the second stage utilizes this estimated distortion information to deblur images and regenerate lost details. Qualitative and quantitative evaluations on both synthetic and real image datasets demonstrate the superior performance of DR-Net over traditional methods, showcasing its broader applicability and robustness. Our approach restores underwater images from a single frame, improving the retrieval of marine resources and enhancing seabed exploration capabilities.
The two-stage architecture of our DR-Net: 1) distorted information estimation network (DE-Net): estimation of distortion/perturbation information from the degraded underwater images, and 2) underwater image restoration network based-GAN (IR-GAN): restoration of image details and reconstruction of potentially lost information.
