This paper presents a novel computational approach for character recognition and grouping of ancient Korean movable metal types printed during the Joseon dynasty. Existing comparative analysis methods lack consistent discrimination between typeface variations in digitized images of characters from historical printed works. The proposed approach applies a three-step process: skeletonization extracts structural features, registration aligns images, and similarity metric computation assesses matching between typefaces. Various preprocessing techniques and metrics are experimentally evaluated. The Dice coefficient emerges as the most reliable discriminative metric. An algorithm is developed employing binarization, skeletonization, iterative closest point registration, and distance computation. Compared to conventional methods, the approach achieves 45 times faster computation while improving character grouping accuracy, enabling more accurate reconstruction of the scale and types of metal letters produced. This methodology provides new insights into historic printing technologies and cultural developments through movable type.