Zone Segmentation and Thinning based Algorithm for Segmentation of Devnagari Text

Abstract

Character segmentation of handwritten documents is an challenging research topic due to its diverse application environment.OCR can be used for automated processing and handling of forms, old corrupted reports, bank cheques, postal codes and structures. Now Segmentation of a word into characters is one of the major challenge in optical character recognition. This is even more challenging when we segment characters in an offline handwritten document and the next hurdle is presence of broken ,touching and overlapped characters in devnagari script. So, in this paper we have introduced an algorithm that will segment both broken as well as touching characters in devnagari script. Now to segment these characters the algorithm uses both zone segmentation and thinning based techniques. We have used 85 words each for isolated, broken, touching and both broken as well as touching characters individually. Results achieved while segmentation of broken as well as touching are 96.2 % on an average.