Underwater Image Enhancement Based on Generative Adversarial Network with Multi-Attention

Underwater image enhancement is a crucial technology for supporting underwater exploration and rescue operations. With the continuous development of the ocean, underwater activities such as archaeology, pipeline inspection, and dredging have been increasing. Nevertheless, owing to the unique complexity of the underwater environment, captured underwater images often face the challenge of information loss, manifested as blurry images, low signal - to - noise ratios, and poor color quality. These issues significantly limit the effective utilization and value of high - quality underwater images in scientific research and engineering practices. Currently, common problems in the field of underwater image enhancement research include high algorithm complexity and insufficient feature extraction capabilities. Therefore, to address the problems of insufficient feature extraction of the target information under test, we adopted the generative adversarial network (GAN) as the main architecture and incorporated an attention mechanism into the network generator. This enables us to centre around the key information of the image during feature extraction, resulting in clearer images. Extensive experiments demonstrate that our method outperforms existing underwater image enhancement methods, validating the effectiveness and the proposed approach.