Character segmentation for multi lingual Indic and Roman scripts

Character segmentation has long been a critical area of the Optical Character Recognition. In this paper, we present an algorithm for character segmentation for Indic and Roman scripts. Character segmentation is difficult for Indic scripts because in these scripts characters are connected with the Shirorekha or headline and the regions bounding the two consecutive characters might overlap because of matraas. Horizontal projection profile is used to extract the Shirorekha and a vertical projection profile is used to segment the characters. Results of the algorithm for the scanned and facsimile documents for Devnagari Script are shown.