This study proposes an underwater image enhancement algorithm based on the pyramid attention mechanism and generative adversarial network (GAN) to improve the enhancement effect of underwater images. It uses the generative adversarial network as the basic architecture, and the generative network adopts the encoding and decoding structures and introduces the feature pyramid attention module. The combination of multi-scale pyramid features and attention mechanism can capture richer advanced features to improve model performance, and the structure of the discriminant network is similar to the Markov discriminator. In addition, a multi-loss function including global similarity, content perception, and color perception is constructed to keep the structure, content,and color of the enhanced image consistent with those of the reference image. The experimental results show that the sharpness, color correction,and contrast of underwater images enhanced by the proposed algorithm are improved. The average values of the structural similarity, underwater image quality measurement, and information entropy are 0.7418, 2.9457, and 4.6925, respectively. For subjective perception and objective evaluation indicators, the experimental results of the proposed algorithm are better than that of the comparison algorithm.