Towards a Hybrid System for the Identification of Arabic and Latin Scripts in Printed and Handwritten Natures

This paper deals with the problem of script identification. Our aim is to use a global approach as a first step towards a hybrid one in order to resolve the problem of differentiation between the Arabic and the Latin scripts in printed and handwritten natures. In our proposed system, features are extracted using three tools namely, Gabor filters, Gray-Level Co-occurrence Matrices (GLCMs) and wavelets. The highest correct identification rate is obtained with GLCMs (84.75%). This result could be assessed as satisfactory when considering the simplicity of the approach and the complexity of the treated problem.