Underwater imagery serves as a vital medium for humans to explore and understand marine environments. Since underwater images suffer from attenuation, color distortion and noise, we propose a style transfer-based framework STUIE-Net to enhance underwater images. STUIE-Net comprises two core components: a preprocessing module and an image enhancement module. The preprocessing module executes red channel prior correction and adaptive style selection, while the enhancement module achieves color restoration through style transfer optimization. Concretely, we develop a red channel prior enhancement algorithm based on color channel mean consistency to rectify color imbalance. Furthermore, we introduce a feature matching-based style selection module that automatically identifies optimal reference styles, effectively resolving the critical challenge of non-reference style transfer. Extensive evaluations demonstrate that our framework significantly improves image clarity and contrast while maintaining natural color distributions. Through both visual comparisons and quantitative metrics, STUIE-Net outperforms existing methods in restoring visually satisfactory results and preserving structural details.