Research on Underwater Image Enhancement Algorithm Based on SRGAN

Abstract

Due to the limitation of the special underwater imaging environment, underwater images usually have problems such as low contrast, blurred texture features, color distortion and so on. Based on the typical problem of underwater images, this paper improves the network structure and loss function on the basis of the original SRGAN network model, and achieves good results. The generative network reduces the convolutional layers and removes the normalization layer (BN layer), reducing resource consumption. The loss function introduces L1 content loss and VGG19 perceptual loss to improve the stability of training. The experimental results show that the improved SRGAN network model effectively solves the color distortion and blurring of underwater images, and has a good enhancement effect on underwater images.