Abstract: The comparative analysis of seals has great significance in the fields of historical research and character relationship. Traditional seal analysis relies on long-term accumulated domain knowledge of experts, it is inefficient and labor-intensive to manually compare seals from a wide range of sources. However, there still lacks of accurate and effective methods. To solve these problems, this paper cooperates with experts in the humanities field to propose an automatic extraction and inpainting method of seals with high-precision, constructs seal portraits from information such as the overall, partial, associated features, and combines intelligent computing with domain knowledge through a visual analysis system. Integrating and presenting a single seal from multiple perspectives helps domain experts conduct an analysis of a seal in a large amount of seal data, thereby greatly improving the work efficiency. This paper validates the method through two cases, and received encouraging feedback from follow-up interviews with experts.