Challenges in segmentation of text in handwritten Gurmukhi script

Abstract

The scanned image of the text is not of any use for user, because that image is not editable. One can not make any change if required to the scanned document. This provides a food for thought for the theory of optical character recognition (OCR). OCR is nothing but character recognition of a segmented part of the scanned image. Therefore the segmented part of the image would be such that it should provide a close relation to the character to be recognised. Hence segmentation plays an important role in the OCR process. There are problems in segmentation process, but the degree of the problems varies from script to script, that is, the problem set for segmentation of the text written in a particular script may differ than the problem set for the text written in other scripts. The characteristics of the script, plays a significant role in deciding the segmentation points. The present study is an effort to find these problems especially for the segmentation of text in Gurmukhi scripts - typed and handwritten. © 2010 Springer-Verlag Berlin Heidelberg.