Devanagari handwritten numeral identification based on Zernike moments

Abstract

The preprocessing of numerals includes bounding them for translation invariance followed by normalization for scale invariance. We achieve translation and scale invariance using simple geometric moments. Higher order Zernike moments are used as shape descriptors. Due to rotation invariance and orthogonal properties of Zernike moments, they are found to perform better in terms of computational complexity and classification achieved. The algorithm has been tested on different handwritten samples taken from different people. In this paper, an attempt is made to develop off-line recognition system for isolated Devanagari numerals.