Abstract: Underwater image enhancement techniques are crucial for the development of marine resources. However, existing underwater image enhancement methods struggle to deal with the complex realities of underwater environments, especially in effectively enhancing challenging samples. To address these issues, we propose a novel method that leverages dual-domain improvements in content and style. Initially, a straightforward and effective framework is designed to decouple underwater images into content features and style codes, followed by reconstruction. For the content domain, a dual-domain adaptive high-frequency adjustment block is designed to enhance detail and suppress noise. For the style domain, we establish a style library that enables style transformation by matching the optimal style during inference. Finally, the model completes the enhancement and reconstruction of underwater images by utilizing a content consistency loss and a reconstruction loss for supervision. Experimental results on publicly available underwater image datasets such as UIEB, EUVP, and RUIE show that our method outperforms existing methods in both subjective visual effects and objective evaluation metrics (PSNR, SSIM, UIQM, and UCIQE). Especially, it demonstrates significant advantages in enhancing challenging samples.