Solving Problems of Thinned Handwritten Arabic words by Zhang and Suen Algorithms

Abstract

The aim of this study is to solve the artificial loops problem resulting from the Thinning process on Arabic names handwriting. The main idea of the proposed solution is to find a threshold for the artificial loop radius this threshold is used to differentiate artificial loops from real ones. The proposed solution has been applied on a suitable and representative subset of SUST- ARG names dataset. To find this threshold opening process has been applied with different radius values on binary images. Those values ranging between 1 pixel and line width of the word where the line width has been estimated using a proposed algorithm, then Zhang and Suen thinning algorithm has been applied. The thesis experiments find that using radius value of 2 pixels gives best result in term of solving artificial loops as well as identify very small numbers of real loops as artificial loops.