An innovative character recognition for ancient book and archival materials: A segmentation and self-learning based approach

Abstract

The paper illustrates the invention of a method and an apparatus able to recognize the text in a set of digital images referring to pages of ancient manuscripts or printed books. It includes the following macro steps: identifying and connecting in sequence regions containing words in a subset of the images; structuring a thesaurus of fonts used in those regions; performing the character recognition of one or more images belonging to the set, associating to this recognition a first value of efficiency. The prototype is patent pending (National Pat. Pend. n. BA2011A000038 - Intern. Pat. Pend. n. I116-PCT). © 2013 Springer-Verlag.