Oracle bone inscription image restoration via glyph extraction

Abstract Oracle bone inscription (OBI) rubbings record the shape, structure, and details of the inscriptions, serving as essential resources for OBI research and digital historical libraries. However, OBIs often suffer from degradation issues such as cracks, erosion, noise, and interference patterns. Existing restoration methods struggle to handle the complex noise in real-world scenarios, frequently leading to glyph distortion or insufficient noise removal. In this paper, we propose a glyph extraction-driven image generation network for high-precision OBI restoration. Our method leverages character glyphs as supplementary information to address complex degradations while preserving the original glyph structures. Additionally, a multi-scale feature extraction strategy is introduced to process noise at varying scales. Comparative evaluations on multiple real-world datasets demonstrate that our method significantly outperforms state-of-the-art approaches, thereby validating its effectiveness for practical OBI image restoration.