The unique multi-cultural, -religious, -ethnic, and -lingual history of the island Elephantine in the river Nile is being studied by texts found on a large body of papyri and ostraca (ERC grant “Elephantine” to V. Lepper, SPK). They cover its history over almost 4000 years. However, many of the papyri are in a very fragile state. Additionally they are often rolled, folded, or fragmented. Manually opening the fragile rolled and folded scripts may severely damage them. On the other hand, a manual assemblage of the large number of fragments (several tens of thousands) is almost impossible. Hence, in these feasibility studies, we investigate some challenges that are related to those precious objects: (i) virtual unrolling & unfolding, based on tomographic reconstruction; (ii) virtual defragmentation, based on photographs; (iii) writer identification of scripts.

 

Virtual Unrolling & Unfolding

In a continuation of a previous feasibility study (see previous project) improvements of the virtual unrolling and unfolding of tomographic reconstructions of papyri are investigated.

Virtual Defragmentation

Goal of the defragmentation is the restoration of ancient documents from fragments by identification and alignment of originally adjacent pieces. Many fragments are rather small and show only little or partial writing. Additionally, neither the total number of original documents nor the extent of missing parts is known.
Sample written papyrus fragments
Sample of written papyrus fragments


Computational methods are studied and developed to support the, currently exclusively manual, puzzling process. Starting point for a computer-assisted defragmentation is the development of a fragment signature that allows for a reliable matching.
Such signatures might include, e.g., a quantification of the ubiquitous papyrus fiber pattern, information about the writing style or fragment shape information.
Papyrus fragment before and after preprocessing with a derived fiber pattern
Papyrus patch before and after preprocessing with horizontal and vertical fiber distribution.


Writer Identification

Current approaches to capture and extract distinguishing features of ancient handwriting are investigated. The identification of scribes might serve the chronological classification of documents and might greatly assist the defragmentation process.