Analyzing Documents in 3D GigaMesh Software Framework

The main focus of the emerging field of Digital Humanities are documents, which are essential to understand the origins of civilization, legislation, religion and history in general. The earliest documents are so-called cuneiform tablets, which are made from clay have wedge-shaped indentations. These cunei (Latin for wedges) are the radical element of a writing system used for several millennia in the ancient Middle East. The hundreds of thousands of texts written in cuneiform script is comparable with the amount of documents in Latin or ancient Greek. Manually drawing and transcribing these tablets is a laborious and tedious task and assistance by an automated and computerized system is highly demanded.

The aim of this project is the extraction of 3D-shapes with high variability, because cuneiform script is handwritten. The crucial steps for feature extraction from 2D-manifolds in 3D-space are reliable edge detection and segmentation. This can be achieved using e.g. integral invariant filtering, a robust technique known from signal processing and shape matching in 2D-space. In order to precisely determine 3D-shapes of characters the pipeline known from image processing and pattern recognition is adapted for optically acquired 3D-models. Millions of vertices approximate an objects surface with an irregular triangular mesh. Different types of integral invariant filtering in multiple scales lead to different high-dimensional feature spaces. Convolutions and combined metrics are applied to the feature spaces to determine connected components i.e. characters. The result is a minimal meaningful vector representation, which is an important basis for paleography and character recognition.

The embedment of the proposed methods in the layered GigaMesh software framework enables a wide variety of applications, which go far beyond cuneiform script. Examples are Roman and Greek inscriptions, medieval seals, fingerprints on ceramics and stripe patterns of papyri.

Link to project website.

Name and contact of project responsible(s):

Dr. Hubert Mara (Interdisciplinary Center for Scientific Computing, Heidelberg University)

Partners

Prof. Dr. Dr. h. c. Stefan M. Maul (Ordinarius der Assyriologie, Uni Heidelberg)
Prof. Dr. Gerfrid G.W. Müller, (Assyriologie, Uni Würzburg)
Dr. Francesco Roberg (Head of Charters Department, Public records office Hessen, Marburg)

Heidelberg Center for Cultural Heritage

Publications:

Vectorization of 3D-Characters by Integral Invariant Filtering of High-Resolution Triangular Meshes
Hubert Mara and Susanne Krömker
Proc. of 12. Int. Conference on Document Analysis and Recognition
(ICDAR/IAPR), pp. 62–66
Washington, DC, USA, 2013

Präzise Bestimmung von Materialstrukturen bei Papyri mit 3D-Messtechnik
Hubert Mara und Patrick Sänger
Zeitschrift für Papyrologie und Epigraphik (ZPE), Band185, pp. 195–199
Hrsg: Werner Eck et. al.
Köln, Deutschland, 2013

GigaMesh and Gilgamesh – 3D Multiscale Integral Invariant Cuneiform Character Extraction
Hubert Mara, Susanne Krömker, Stefan Jakob and Bernd Breuckmann
Proc. VAST Int. Symposium on Virtual Reality, Archaeology and Cultural
Heritage, pp. 131-138
Palais du Louvre, Paris, France, 2010