Script Identification from Printed Document Images Using Statistical Features

Automatic identification of a script in a document image facilitates many important applications such as automatic archiving of multilingual documents; searching online archives of document images and for the selection of script specific OCR in a multilingual environment. In this work a technique for script identification from document images is proposed. The method uses vertical and horizontal run components/objects of words of a single line of text to distinguish 3 Indian scripts: Kannada, Hindi and English. Initially, the method segments words from the selected line of text from a document image. Then statistics of horizontal and vertical run objects are determined. Further, linear discriminant function is used to identify script of the document image as Kannada, Hindi or English script. The method has been tested for 300 document images and the method found to be robust and efficient. The proposed system achieves 93% identification accuracy for Hindi script, 90% identification accuracy for English script and 86% identification accuracy for Kannada script.