Text and graphics segmentation of newspapers printed in Gurmukhi script: a hybrid approach

Abstract

Newspapers are always a standard medium to convey important information to masses of people in recent time as well as in old time. An automated system is required to convert information into a processable form so that information could be searchable. Many efforts have been done on Gurmukhi script documents in typed or written form, but very few articles are present on Gurmukhi script newspaper text recognition or text and image segmentation. Image/graphics segmentation of text is mandatory before feeding newspaper text to OCR for accurate results. In the literature, many techniques have been proposed for segmenting images and text, but many are complex in nature. In this article, the authors have proposed a very simple and effective hybrid approach based on run length smoothing algorithm and projection profile to segment an image from text in Gurmukhi script newspaper articles. Both horizontal and vertical run length smearing algorithm is used for labeling the regions. Logical AND operator is applied to resultant images to identify the text and image regions. To segment the image region among the labeled regions, projection profile technique is implemented. The combination of these two techniques has produced very good results.