Identification of Devanagari Script from Bilingual Printed Text Documents

Abstract

Bilingual script identification is one of the challenging steps in the development of optical character recognition (OCR). As India is a multilingual multiscript country, in its constitution for each state, the respective state language and the script is used for state official work. In Maharashtra, Marathi is the state official language with Devanagari script and English as the communication language. We need to develop the OCR that can identify and differentiate both scripts. This paper presents a research work for identification of Devanagari (Marathi) script from printed bilingual text document. In our work, we have developed a methodology that applies projection profile in line segmentation which is followed by twofold word segmentation. We examined the use of structural features (header-line pixel count and intercharacter gap) as a tool for determining the Devanagari words. The heuristic rule approach is used for classification. The proposed method is implemented on ten printed bilingual text images. These images consist of 77 lines and 474 words of varying font sizes. The result of our experimentation shows accuracy of 87.25 % in identification of Marathi words.