Abstract:
The comparative analysis of seals has great significance in historical research, cultural archaeology, and other fields. Traditional seal analysis relies on the long-term accumulation of domain knowledge by experts and requires manual comparison among widely sourced seals, which is inefficient and extremely labor-intensive. To address this problem, we collaborated closely with experts in the field of humanities to propose an automatic extraction and restoration method for ancient painting seals, achieving high-precision extraction and restoration of complete seals. A dataset of 1 762 ancient painting artworks and their seals obtained through the internet is used to construct seal portraits based on information such as the overall, local, and associated features of the seals. A visual analysis system was designed to integrate and present seals from multiple perspectives, helping domain experts to compare and analyze specific seals within a large dataset. Compared with traditional manual analysis, this method greatly improves the work efficiency of experts. Two case studies have confirmed the effectiveness of the proposed method and system, and received positive feedback in expert interviews.