المستخلص: |
In this paper, we propose a method of an Arabic and Latin discrimination writings, based on morphological, geometrical and statistical analysis. We focused on feature extractions in order to curry out a system able to differentiate between printed or handwritten Arabic and Latin scripts. First, an image acquisition is made by means of a scanner, and then come some pre-treatment steps. After that, we extract the different features constituting an input vector of multi¬layer perceptron (MLP) with back-propagation algorithm. Finally and thanks to the network outputs we have been able to undertake a discrimination of each script (Printed or handwritten). The aim of our work is to constitute a postal sorting system in French-Arabic sneaking countries.
|