A new hybrid binarization method based on Kmeans

Mahmoud Soua (1) , Rostom Kachouri (1) , Mohamed Akil (1)

1 LIGM - Laboratoire d'Informatique Gaspard-Monge

The document binarization is a fundamental processing step toward Optical Character Recognition (OCR). It aims to separate the foreground text from the document background. In this article, we propose a novel binarization technique combining local and global approaches using the clustering algorithm Kmeans. The proposed Hybrid Binarization, based on Kmeans (HBK), performs a robust binarization on scanned documents. According to several experiments, we demonstrate that the HBK method improves the binarization quality while minimizing the amount of distortion. Moreover, it out-performs several well-known state of the art methods in the OCR evaluation.

Mots clés: Scanned documents; binarization; Kmeans; OCR

Domaines: Informatique [cs]; Traitement du texte et du document; Traitement du signal et de l'image [eess.SP]

Mahmoud Soua, Rostom Kachouri, Mohamed Akil. A new hybrid binarization method based on Kmeans. 6th International Symposium on Communications, Control and Signal Processing (ISCCSP), May 2014, Athens, Greece. ⟨10.1109/ISCCSP.2014.6877830⟩. ⟨hal-01305856⟩