Entropy based Script Identification of a multilingual Document Image

Abstract

Automatic Document Image Analysis has been a prime field of research in the past few decades. Script Identification is an essential part of automatic document image analysis. Script is essentially the text of a written document and languages are written using them. A huge set of techniques have been proposed and many scripts, foreign & domestic, have been identified. But so far, trivial work has been reported for the identification of Kashmiri script. In this paper we are proposing & experimentally testing identification of Kashmiri script collectively with three other related scripts viz. Roman, Devanagri & Urdu using entropy. First, a set of training images are experimented to prepare the knowledge base and later the actual samples have been evaluated. The proposed system offers an accuracy rate of 98.50%.