Adaptive Structural-Frequency Modeling for Mural Image Restoration
Fanhua Zhao, Hui Ren · 2025
Ancient murals have suffered from cracks, peeling, and fading due to prolonged natural erosion and human impact, diminishing their artistic value and historical significance. To address the issues of detail loss and structural inconsistency in mural restoration, this study proposes a novel Adaptive Structural-Frequency Modeling (ASFM) approach. ASFM integrates feature modeling and frequency optimization, enhancing structural coherence in the spatial domain while preserving fine-grained texture details in the frequency domain. Experiments conducted on the Dunhuang mural dataset demonstrate that the proposed method outperforms state-of-the-art approaches in terms of PSNR, SSIM, and LPIPS, while achieving more visually natural restoration results. Ablation studies further verify the contributions of the feature modeling and frequency optimization modules, highlighting their effectiveness in improving both overall consistency and local detail restoration. This research provides an effective solution for computer-aided mural restoration and can be extended to other digital preservation tasks for cultural heritage.