Issue No. 08 - August (2009 vol. 31)
Dimitra Dafi , National Technical University of Athens, Athens
Panayiotis Rousopoulos , National Techncal University of Athens, Athens
Stephen Tracy , American School of Classical Studies at Athens, Athens
Michail Panagopoulos , National Technical University of Athens, Athens
Constantin Papaodysseus , National Technical University of Athens, Athens
This paper introduces a novel methodology for the classification of ancient Greek inscriptions according to the writer who carved them. Inscription writer identification is crucial for dating the written content, which in turn is of fundamental importance in the sciences of history and archaeology. To achieve this, we first compute an ideal or "platonic” prototype for the letters of each inscription separately. Next, statistical criteria are introduced to reject the hypothesis that two inscriptions are carved by the same writer. In this way, we can determine the number of distinct writers who carved a given ensemble of inscriptions. Next, maximum likelihood considerations are employed to attribute all inscriptions in the collection to the respective writers. The method has been applied to 24 Ancient Athenian inscriptions and attributed these inscriptions to six different identified hands in full accordance with expert epigraphists' opinions.
Writer identification, handwriting classification, ancient Greek inscription classification, handwriting analysis, archaeology, feature modeling, pattern recognition.
Dimitra Dafi, Panayiotis Rousopoulos, Stephen Tracy, Michail Panagopoulos, Constantin Papaodysseus, "Automatic Writer Identification of Ancient Greek Inscriptions", IEEE Transactions on Pattern Analysis & Machine Intelligence, vol. 31, no. , pp. 1404-1414, August 2009, doi:10.1109/TPAMI.2008.201