Two-stage underwater image restoration based on gan and optical model

Due to the unique characteristics of the underwater environment, the underwater images often have the problems of blurring and hazing, which affects the identification of image details. To enhance the details of the image, this paper proposes a two-stage restoration method based on the underwater optical model and generative adversarial network. Firstly, we synthesize the paired datasets using the underwater imaging optical model. Then, a two-stage deep learning method is employed to process the underwater images. In the first stage, the images are dehazed; in the second stage, the details of the image are improved. Finally, quantitative and qualitative experiments were conducted to evaluate the performance of the proposed method. The qualitative results show that compared with other state-of-the-art methods, our method can better highlight the image details and effectively improve the visual effects of the images. In the quantitative evaluation, the images restored using the method proposed in this paper achieved higher scores in each of the metrics, which proves to the effectiveness of the proposed method.