Research on hydrological image restoration based on LGNet

There are a large number of literature materials about the Shui nationality, but these materials are badly damaged and difficult to identify, and the traditional restoration methods have limitations in time and space, which affects the inheritance of the Shui nationality culture. In this paper, a generative adversarial network of global Discriminator and local discriminator is designed. The local discriminator is used to reconstruct the restricted discriminator well, and the global discriminator ensures that the overall semantic information of the image remains unchanged. At the same time, residual links are built at each stage of the generated network, avoiding the problem of disappearing gradients and allowing deeper networks to learn effective features. This model provides an efficient, accurate and automatic method for the restoration of Shui script, and effectively speeds up the restoration process of Shui ancient characters.