South Indian handwritten script identification at block level from trilingual script document based on Gabor features

Abstract

The script is a graphical illustration of thinking of a person. Any script can be considered as texture patterns which have linear, oriented and curvilinear sub-pattern primitives. In this paper, the problem of automatic handwritten script identification is considered as texture analysis problem. This paper presents the significance of the traditional Gabor filters in extracting oriented energy distributions. These are tuned efficiently with 24 channels to extract directional energies of text blocks of each script. K nearest neighbor classifier is employed for discriminating six south Indian scripts based on the standard deviations of Gabor filters response. The comprehensive experimentation is conducted on a data set of 600 text block images. Average tri-script classification accuracy with two fold cross validation is 91.99 %. © 2013 Springer.