Paper Details: Downloads: 1713
Serial Number: P1151549451
Title: Text lines Segmentation of Handwritten Arabic Script using Outer Isothetic Cover
Authors: Samia Snoussi and Fethi Ghazouani and Yosra Wahabi
Abstract: In this paper, we present a new bottom up segmentation method of handwritten Arabic texts into text lines. The proposed method uses the algorithm for construction of the Outer Isothetic Cover (OIC) of a digital object. This technique allows the construction of polygons of all connected components on the binary image of the document. The specificity of this method is that it allows the extraction of shape of the script and not a line as a geometric form. This will reduce overlapping between successive lines. The second specificity of this method is its independency from the variability of script size in the same document. The set of extracted polygons contains big and small ones. The big polygons correspond to Pieces of Words and the small ones correspond generally to diacritic points or punctuation mark and they are identified by their perimeters after a training step. Based on these results and added information like position, upper and lower limit of the components given by the OIC, we can extract lines from handwritten Arabic text. The proposed method is tested and evaluated on a sub-set of 100 randomly chosen handwritten Arabic texts selected from KHATT Database. The obtained result attempts 74% and needs to be improved. A new database composed by more than 1000 pages of printed and handwritten documents for tutorial and guiding of HAJJ steps is created (The Hajj is the pilgrimage to Mecca). We are working on the application of the proposed segmentation method on the HAJJ document database in order to extract desired information.
Keywords: text Segmentation, handwriting, Outer Isothetic Cover, KHATT Database, HAJJ Database.
Journal/Conference: International Journal of Graphics, Vision and Image Processing
Volume: 16
Issue: 1
Submission Date: 12/2/2015 12:00:00 AM
Review Date: 12/23/2015 12:00:00 AM
Publishing Date: 2/12/2016 12:00:00 AM
Article Downloads: 1713
Download:

Facebook