Handwritten Arabic Digits Recognition Using Bezier Curves
In this paper we propose a new recognition approach for Arabic numerals. Given that the performance of recognition systems for Arabic numerals are closely linked to the choice of features and classification system used in the recognition phase, we seek to exploit the possibilities of the theory of Bzier curves that allows representing parametric curves from a limited number of data (some characteristic dots with their derivatives). Indeed, the characteristic dots of the Arabic digit that we have adopted are those such that their associated Bezier curve is close to the shape of the digit. The used classifier in this work is the k-nearest neighbour. The obtained results testify to the interest and the strength of our approach.
Keywords: Handwritten digits recognition, Image processing, Spline, Bezier curves, Skeletonization, Feature extraction, Training.
Download Full-Text
ABOUT THE AUTHORS
Aissa Kerkour El Miad
•PhD student .Member of Laboratory of Researches in Computer Sciences (LaRI) •Since 2007, he prepares his PhD thesis in Recognition of handwritten and printed Arabic characters and digits within the laboratory LaRI. •His research interests are especially in developing OCR for Arabic language.
Azzeddine Mazroui
•Doctorat d’Etat in numerical analysis, Mohammed First University Morocco, 2000 •PhD in probability and statistics, Pierre & Marie Curie University France, 1993 •Full professor of mathematics and computer sciences in University Mohammed First •Member of Laboratory of Researches in Computer Sciences (LaRI) •Director of the ANLP unit in the LaRI laboratory •Cofounder of the UFR Doctorate in "Approximation, Computer Sciences and Signal Analysis".
Aissa Kerkour El Miad
•PhD student .Member of Laboratory of Researches in Computer Sciences (LaRI) •Since 2007, he prepares his PhD thesis in Recognition of handwritten and printed Arabic characters and digits within the laboratory LaRI. •His research interests are especially in developing OCR for Arabic language.
Azzeddine Mazroui
•Doctorat d’Etat in numerical analysis, Mohammed First University Morocco, 2000 •PhD in probability and statistics, Pierre & Marie Curie University France, 1993 •Full professor of mathematics and computer sciences in University Mohammed First •Member of Laboratory of Researches in Computer Sciences (LaRI) •Director of the ANLP unit in the LaRI laboratory •Cofounder of the UFR Doctorate in "Approximation, Computer Sciences and Signal Analysis".