Identification of handwritten text in machine printed document images

Abstract

In our daily lives we come across many documents where both printed and handwritten text co-exist and sometimes intermingle. As the OCR techniques for processing the two are quite different it is necessary to classify and distinguish them first. In this paper, a scheme has been proposed by which handwritten, printed and "mixed" text regions in the same document image can be identified and demarcated from each other for Bangla, the second most popular Indian script. The proposed scheme has been established on the basis of the structural and statistical idiosyncrasies of printed and handwritten Bangla text. © 2013 Springer-Verlag.