In recent years, the application of deep learning in image generation, visual recognition and creative design has deepened, and neural style transfer has gradually become an important technical path connecting computer vision and art design. Aiming at the problems of aging visual expression, limited style update and easy weakening of cultural recognition of traditional stone lion images in contemporary communication scenes, this paper constructs a redesign method of stone lion images based on neural style transfer, and proposes a computational aesthetics framework for cultural symbol modernization. The multi-source data organization is carried out around the traditional stone lion images, modern art style samples and visual design samples. Through image preprocessing, deep feature representation, bidirectional style mapping and aesthetic constraint modeling, the coordination and fusion between the content structure of the stone lion and the modern visual style are realized. The experimental results show that the proposed method can effectively enhance the contemporary expression features of color, texture and composition while maintaining the head contour, mane hierarchy and overall pose recognition of the Shi lion. In the later stage of model training, the cycle consistency loss is reduced to 1.3, the average PSNR reaches 25.33 dB, and the average SSIM reaches 0.844. The results show that the proposed method has good application value in the field of traditional cultural image digital updating and visual redesign.
