Improved Scale-Invariant Moments for Deformation Digits

Abstract

This paper presents improved scaled-invariant moments for digits with deformations. We claim that deformation digits would be digits with improper shapes, unconstrained styles of writing and different orientations. An experimental evaluation of utilizing various moments order as pattern features in recognition of handprinted and handwritten digits have been carried out using improved scaled-invariant moments. We use scale-invariant moments of order 2 for the numerator and order 4 for the denominator while preserving the scale factor of the same order. These moments have been used as feature extraction for digits with various orientations. Digits are rotated clockwise and counter clockwise of 45 degree, and using unequal scaling in x and y directions. As a comparison, we generate geometric moment invariants on handwritten digits. We train improved scaled-invariant moments using a standard backpropagation model and modified backpropagation model for classifications. We found that the results are promising with improved scaled-invariant moments of higher order, and the classifications of digits are successfully recognized.

Keywords:

• Handwritten digits
• handprinted digits
• feature extraction
• geometric moment invariants
• backpropagation model

C. R. Categories::

• I.5.1
• I.5.2