Image Generation Framework for Digital Intangible Cultural Heritage Pattern Reconstruction Using GAN Fusion Shape Grammar
Y. Qiu, T. Xue · Advanced Electromagnetics · 2026
Digital reconstruction of traditional intangible cultural heritage patterns remains challenging because of their complex geometric structures, rich cultural semantics, and the need to balance authenticity with creative generation. This study proposes an image generation framework that integrates shape grammar with conditional generative adversarial networks (cGANs) to achieve high-quality reconstruction of traditional patterns. Shape grammar is first employed to define geometric composition rules and structural constraints, which are then embedded into the cGAN as conditional information to guide pattern generation while preserving cultural characteristics. The discriminator further optimizes the generation process through adversarial learning to improve image quality and structural fidelity. With the increasing deployment of intelligent visual sensing, wireless information acquisition, and digital communication systems, accurate reconstruction of cultural pattern information provides valuable support for multimodal data transmission and digital content representation in electromagnetic-enabled smart environments. Experimental results demonstrate that the proposed framework achieves style consistency exceeding 0.75 and cultural fidelity scores above 3.2 (on a five-point scale), outperforming conventional generation methods in both reconstruction accuracy and creativity. The proposed framework offers an effective solution for digital preservation and intelligent innovation of intangible cultural heritage and provides methodological references for visual information processing in future electromagnetic sensing and communication applications.