The Effective Use of the Histogram of Principal Oriented Gradients for Natural Arabic Image Character Recognition

Nowdays, recognizing characters from natural scene image is an important task in various applications of pattern recognition. In fact, the automatic recognition of characters from natural scenes allows providing many information for peoples such as the language translation from smartphone or the address identification from a camera transported in a vehicle. Hence, most of the systems are implemented for recognizing the English language by employing different robust descriptors and classifiers. However, few works are dedicated for the Arabic characters. Also, this paper aims to focus on the use of a recent descriptor namely the Histogram of Principal Oriented Gradients (HPOG) that is used in our best knowledge for the first time in character recognition. For classification, the One Class-Principal Component Analysis (OC-PCA) Classifier is used for recognizing Arabic characters in natural scene. For evaluating the performance of the HPOG associated to the OC-PCA, experimental results conducted on a standard Arabic dataset show the effectiveness of the proposed system against the state-of-art.