Restoration of Underwater Distorted Image Sequence Based on Generative Adversarial Network

Underwater images will be distorted due to the influence of turbulence, and images will appear geometric distortion since the light is refracted by the turbulence, which makes task of image recognition difficult. In order to improve image recognition underwater, this paper proposes an image restoration method using underwater distorted image sequence through deep learning technique. Considering the complexity of dynamics motion, image sequence is more feasible to realize task of restoration, which contains enough information of water turbulence. Generative adversarial network as a deep neural network has proved to be an appropriate method applying to the field of image processing, which is used to restore the distorted image. Experiment shows the proposed method has fine ability of using distorted image sequence to realize image restoration.