Automatic characterization of the border deterioration in Byzantine seals

Byzantine seals were attached to official documents to authenticate the sender, thus enclosing a valuable part of the Byzantine Empire’s history. With the aim to provide computational models to help historians in the seal interpretation, this paper proposes a method to automatically determine the level of deterioration of the seal borders from their photographs, i.e., a non-destructive inspection. The method consists of a segmentation step based on Morphological Geodesic Active Contours, a feature extraction step, and a classification step that groups the seals into four border deterioration categories. Our best results reached 0.80 in accuracy, 0.73 in Cohen’s Kappa statistics score, and 0.80 in the macro F1 score. • Classifying Byzantine seals by their border damage is useful since most are broken. • Byzantine seal characteristics indicating border deterioration level are analyzed. • The first border deterioration-based automated Byzantine seal classification. • A method based on segmentation, feature extraction, and classification is proposed. • Accurate classification results are obtained (accuracy = 0.80).