Automatic dewarping of Camera Captured Born-Digital Bangla Document Images

Unless special arrangement is made, the camera captured image of a thick book page gets distorted because of bending of the page. Nowadays students and common people take text images from curved surfaces using smart phone which too gets distorted in a similar manner. The distortion of such born-digital document images should be removed for automatic extraction of various information from such images. The techniques proposed in the literature for such task mostly work on alphabetic script like English. We did not come across any published work on alpha-syllabary (e.g. Indic) or mixed-script document. This paper is an attempt to fill this gap. Also, the proposed approach tries to restore the font shape including size and style (e.g. Italic, Bold) in a better way. Experimental results obtained on various document images have shown promising results.