A Novel Baseline Estimation Method for Arabic Handwritten Text Based on Exploited Components of Voronoi Diagrams

Atallah AL-Shatnawi

Department of Computer Information System, Al-albayt University, Jordan

Abstract: The goal of this paper is to present an efficient novel baseline estimation method for Arabic handwritten text based on the exploited components of Voronoi Diagrams (VD). The proposed based-VD method is constructed from three stages including: Preliminary stages, VD construction and baseline estimation process. The edges of the text are firstly extracted and then both inner and outer contour are traced in order to be converted into a set of sampling points. These sampling points are used to be the VD generators. Then, the baseline is detected from those edges and vertices which are positioned within the text boundaries. The proposed method is implemented, verified and validated on the IFN/ENIT Arabic handwritten dataset. It is discussed and compared with the horizontal projection method against the IFN/ENIT dataset based on affecting by noise, working properly with or without diacritics, working properly with the skewed images and running time efficiently. The results obtained are shown in graphs and output images. The results demonstrated that the proposed method works properly with the skewed and noisy images and with or without diacritics. It also, able to estimate the Arabic text baseline in straight or in curved line.

Keywords: Arabic text recognition, baseline estimation, VD, horizontal projection.

Received December 7, 2013; accepted July 9, 2014