Dating the Historical Documents from Digitalized Books by Orthography Recognition

This paper introduces a new method for automatically dating Serbian and Croatian historical documents. It is based on the concept that the documents in a certain script or language evolving in different historical periods are characterized by differences in orthography rules. Accordingly, we propose three stages of script coding, texture analysis and classification for capturing such a difference. Hence, the input document is transformed into a sequence of numerical codes, each representing an intensity value, determining an image. Then, texture analysis extracts features from the image to create a feature vector. Finally, it is classified for orthography recognition. Results obtained on two databases of historical documents in angular Glagolitic script and Slavonic-Serbian and Serbian languages extracted from digitalized books demonstrate the efficacy of the proposed method.