Objective  This study focuses on addressing the issue of local pattern loss and blurring in Tang Dynasty beaded pattern images caused by the damage of cultural relics or insufficient digitalization accuracy, and explores its application in modern leather handbag design to achieve the effective inheritance and innovative activation of this traditional pattern, providing technical tools for the high-fidelity digital restoration of cultural heritage. Through the integration with modern design, this study explores the application value of traditional patterns in contemporary life and promotes the creative transformation and innovative development of China's fine traditional culture.

Method  The core innovation of the improved SwinIR algorithm proposed in this study lies in the use of a spatial frequency module constructed by fast Fourier convolution to replace the 3×3 convolutional layers in the original network, significantly enhancing the model's ability to fuse global and local features of images. At the same time, the Charbonnier loss function is adopted to replace the traditional L1 and L2 loss functions, which better balances the smoothness of errors and the robustness of the model during the repair process. The algorithms SwinIR and Real-ESRGAN are verified through comparative experiments. The improved algorithm performs the best in visual restoration effect and can restore the local details, yarn texture and original color of the beadedpattern more clearly.

Result  Compared with Real-ESRGAN, the peak signal-to-noise ratio (PSNR) is increased by 0.92 dB, and the structural similarity index (SSIM) is increased by 0.0636. Compared with the original SwinIR, the PSNR is increased by 0.63 dB and the SSIM is increased by 0.0062 dB. The average opinion score (MOS) of the subjective evaluation index is the highest, which is 1.68 points higher than that of Realesrgan and 0.44 points higher than that of SwinIR, respectively, indicating that the quality of the restored image is unanimously recognized by experts. Using high-quality Tang Dynasty beaded pattern images restored by the improved algorithms as design materials, this study applies shape grammar to conduct design methods such as replacement, addition and deletion, scaling, copying, rotation and cross-cutting of pattern elements, as well as deconstruction and reconstruction. Representative beaded patterns are selected and recreated around the auspicious theme of "jade beads strung together, complementing each other perfectly". The design process preserves the core cultural connotation and typical artistic features of the Tang Dynasty beaded pattern, while integrating modern design language. The final design scheme is applied to leather handbags to effectively integrate the cultural charm of traditional patterns with the trendy and fashionable sense of modern bags.

Conclusion  This algorithm provides a powerful tool for the high-fidelity digital restoration of cultural heritage. It provides a practical technical path and design paradigm for the digital protection, dynamic inheritance and innovative application of Tang Dynasty beaded patterns and other traditional designs in contemporary design, especially in leather product design, which has positive significance for promoting the creative transformation and innovative development of China's fine traditional culture.