Document processing methods for Telugu and other South East Asian scripts

It is observed that in several South East Asian scripts, a single character consists of two or more connected components. In these scripts the complex arrangement of connected components leads to problems such as touching characters and difficulty in identifying words and text line boundaries. In the present work we propose a method to extract text lines by clustering of connected components, based upon their spatial properties. Those components, with abnormal properties and which are not identified by an OCR, are sent for character segmentation. For character segmentation we describe "Drop Fall" and "Whitestream" methods. The methods presented here are applicable to any language (script) that requires connected component based processing.