A re-assembling scheme of fragmented Mokkan images

Historical documents are invaluable to study the society and culture in old ages everywhere in the world. In Japan, unearthed wooden tablets called Mokkan excavated from ancient palace sites and so on in the Nara period provide important clues to know the era. Since most of unearthed Mokkan have been badly damaged and broken into several pieces, however, it is extremely difficult even for experts to extract characters on fragmented Mokkan. In this paper, we propose a digital image reassembling scheme for fragmented Mokkan so that broken character images are reassembled and written content is analyzed. The proposed scheme consists of two steps: an image grouping using color features and an image reassembling using local tangent and curvature functions of the fragment contours. After the grouping process, fragment images with the same color features are clustered. Then, in the reassembling step, candidate matching pairs for adjacent fragment images in the same group are listed. We also provide a user interface for archeologists to verify the results. As a result, the system helps archaeologists reconstruct Mokkan images so that they can decode them.