Title: Comparison of vertical projection profile, moment-based and initial skew rate algorithm for text skew estimation
Authors: Darko Brodić; Ivo R. Draganov; Zoran N. Milivojević; Viša Tasić
Addresses: Technical Faculty in Bor, University of Belgrade, Vojske Jugoslavije 12, 19210, Bor, Serbia ' Department of Radio Communications and Video Technologies, Technical University of Sofia, 8 Kliment Ohridski Blvd., 1000, Sofia, Bulgaria ' College of Applied Technical Sciences, Aleksandra Medvedeva 20, 18000, Niš, Serbia ' Department of Computer Science, Mining and Metallurgy Institute, Zeleni bulevar 35, 19210, Bor, Serbia
Abstract: This paper makes analysis and evaluation of vertical projection profile, moment-based and initial skew rate algorithm for the text skew. Among these three methods, the initial skew rate method is newly developed, while moment-based is adapted compared to its initial version. The vertical projection profile method is used in its native form, which is widely accepted for text skew estimation. The comparison is based on the experiment which includes testing of the algorithm with a dataset that consist of the printed text image samples. These image samples are given in the resolution of 25, 50 and 300 dpi. Tested algorithms obtain different skew accuracy for different resolution of text images. At the end, the moment-based method has the smallest accuracy declination, which demonstrates its benefit over the other two methods. Furthermore, this contributes to its robustness in applications, which use low resolution images.
Keywords: binarisation; initial skew rate; ISR; moment-based method; printed text documents; projection profiles methods; text skew.
International Journal of Reasoning-based Intelligent Systems, 2017 Vol.9 No.2, pp.90 - 100
Received: 31 Mar 2016
Accepted: 08 Mar 2017
Published online: 14 Dec 2017 *