Kannada Handwritten Numerals Recognition and Translation Using Template Matching

Abstract

Template matching technique equates portion of input character images versus trained character images, Where Convolution is one of the method in template matching. This convolution technique can be used to recognize images, scanned documents and even handwritten scripts. In this work convolution is used to recognize the hand written numerals of Kannada by using pen, pencil, sketch and machine printed digits. The convolution method is used to differentiate 2000 numerals based on style, stokes, texture, thinness, thickness, darkness and even fonts size. The steps involved in convolution for template matching are Scanning of hand written document to get preprocessed binarized noise free images .in the next step comparison will takes place between input and template images which is totally depend upon correlation between the two. If the correlation is high matching is successful between handwritten written numeral images and trained images. The average set gives out 86.28% matching efficiency. By comparing all the results which is obtained from the work reveals that sketch handwritten is the suitable than pencil and pen in template matching.