In this paper, a knowledge assisted automatic recognition system of sacred figures in Byzantine icons is proposed. The painters of the Byzantine and post Byzantine works of art use specific rules of iconographic patterns for the creation of the sacred figures. Based on these rules, a Byzantine icon is separated into semantic regions, with specific characteristics that can make the sacred figure distinguishable. The proposed system detects and analyzes the most important facial characteristics providing by that way rich but imprecise information about a Byzantine icon.
In order to effectively represent this information a fuzzy knowledge base was defined using the expressive DL f-SHIN. The extracted information comprises the assertional component while an expressive terminology that was defined according to Byzantine Art treatise permits categorization of Byzantine artworks. The proposed approach was tested on a repository of 2000 Byzantine icons presenting very promising results.