Printed Telugu Numeral Recognition based on Structural, Skeleton and Water Reservoir Features

Abstract

Selection of feature extraction method is most important factor in achieving  high  recognition  performance  in  automatic  numeral recognition  systems In this paper, a new proposed system for Telugu printed numeral characters  recognition using number of contours, skeleton feature, water reservoir features, and ratio of length of Top line to bottom line of the image. Printed Telugu numerals are scan converted to binary  images and extract the features irrespective of size of the image. The proposed method is applied  with  success  to  a  database  of  3150  printed  multi-font printed  Telugu numerals  The  recognition  accuracy is achieved 100%. The  experimental  results  obtained  are encouraging  and  comparable  with  other  methods  found  in literature survey. The novelty of the proposed method is free from size normalization and without classification method.