Image style transfer lies at the intersection of computer vision and artistic design, and has been widely applied in various visual generation tasks. However, existing methods still suffer from limitations in generalization, computational efficiency, and structural fidelity. In this work, we propose the Wavelet Transfer Network (WTN), a training-free style transfer method that integrates wavelet transforms into an image reconstruction framework, and explores its application in museum cultural and creative design. Experiments on 120 content–style pairs show that WTN reduces style loss to 6.9, achieves the highest user preference score of 31.2%, and runs at 0.91 s per image, striking a good balance between fidelity and efficiency. The results of various museum artifacts and artistic styles confirm that WTN preserves visual details, restores structural integrity, and ensures consistent stylization. Furthermore, WTN generalizes well to other style transfer tasks and broader artistic domains, without the need for additional training or fine-tuning.