Moment-based extraction on handwritten digits

Abstract

Handwritten digits recognition software have become a highly demand applications to the market. Manufacturing industries as well as post offices are among the users of these applications. In the past few years, several approaches have been used in development of handwritten recognition applications. However, the accuracy of recognition varies between one and another. In this study, the approach of moment-based techniques are employed on handwritten characters.. These include geometric moments, Zernike moments and contour sequence moments. Classification and recognition results are analyzed to determine the necessity of operation thinning when dealing with the moment functions. A Simple Block Segmentation with Moore Tracing Algorithm (SBS & MNTA) is used in image segmentation while Safe-point Thinning Algorithm (SPTA) is applied in image thinning process. Results obtained have shown that operation thinning should be excluded as its deteriorates the recognition accuracy. Contour sequence moments exhibited the highest recognition rate compared to Geometric moments and Zernike moments.