Shape description for Arabic character recognition: A comparative study

Abstract

In this paper a shape description for Arabic character recognition was presented. In the literature, shape description approaches are classified into either statistical or structural. In our work, we proceeded with a comparative study to compare the different shape descriptions using the same approaches and dataset of the Arabic characters. Our finds proved that the structural approach relying on the Freeman code chain achieved better recognition rates. In addition, the combination of geometrical moments, the Fourier Descriptor and the structural feature allow to improve the rates even when compared to the structural approach.