dc.description.abstract | Many books were printed with movable metal types during the Joseon Dynasty (A.D. 1392-1897), and even early books are older than the Gutenberg Bible which was the earliest book printed using movable metal type in Europe (A.D. 1455). However, there is little information about the printing technology, such as the scale of metal type at that time and the management system. In this paper, we propose an algorithm to estimate the number of chases simultaneously used to print a book with movable metal types, which can provide not only insights into the level of printing technology at that time but also important clues for estimating the number of movable metal types. In contrast to previous studies of chase count estimation based on the subjective comparison, we propose an algorithm to estimate the number of chases using the similarity of character spacing distribution in the central part of the printed sheets, assuming that the metal types for the central part were set once and reused without replacement during the production of the book. The central part of each sheet in one book is cropped. After rotational deviations of all the central part images are removed using the principal component analysis, the images are binarized. Morphological operations are performed to remove noise and facilitate character analysis. The profile histogram of the binarized image is used as a probability distribution for character spacing, and the Wasserstein distances among the profile histograms of all the central part images in the book are calculated to obtain a similarity matrix of the central part images. By performing an eigenvalue analysis using spectral clustering with the similarity matrix, we determine the number of clusters of the central part images, which indicates the number of chases simultaneously used in printing the book. When applying the proposed algorithm to the images of "Neungumgyeoung" Volume 4 (A.D. 1462), "Seokbosangjeol" Volume 6 (A.D. 1446), and "Worin cheongangjigok" Volume 1 (A.D. 1447), it was estimated that 3, 6, and 4 chases were used, respectively. To verify the reliability of the results, subjective classification of the core part based on the size or style of the characters was performed for "Neungumgyeoung" Volume 4, and it was classified into 3 groups, consistent with the results of the proposed algorithm. This proposed method can be utilized for the analysis of various metal movable chases. | en_US |