Segmentation of printed bangla characters using structural properties of Bangla script

Some difficulties arise in segmentation of Bangla characters because of their unique structural properties, such as, having connectors adjoining two characters and vowel modifiers modifying either the top portion or the bottom portion of a character. This paper presents some ways to overcome these problems by taking into account these unique properties and thereby increases the accuracy of character segmentation of printed Bangla scripts that leads to greater accuracy of optical character recognition applications.