Fast Technique to Recognize Closed Characters in Arabic Handwriting

Abstract

Considering the distinctive features and characteristics of Arabic scripts, recognizing handwritten Arabic characters is still widely open as a major experimental task, which is not easy to overcome. This study attempts to develop a fast and effective technique for recognizing this difficult type of written script by using an iterative algorithm to extract closed Arabic characters. Some Arabic characters have circles that may be blurred, incomplete, unformed, or distorted. In addition to ligatures with different shapes, the technique is based on the development of a proper Circular Hough Transform. This study used IFN/ENIT, which is a standard database for extracting closed Arabic characters as designed in classes. The first class is closed blurred circles ‫,و(‬ ‫ع‬ , ‫م‬ ‫.)ق,‬ The second class is unformed circles ( ‫ة‬ , ‫ع‬ ‫.)ق,‬ The third one is incomplete circles ‫ض(‬ ‫ص,‬ ‫ظ,‬ , ‫ط‬ ‫.)ه,ة,‬ Finally, the fourth class is distorted circles ( ‫ظ‬ ‫ط,‬ ). This methodology achieved an outstanding accuracy of approximately 96.67%, a very interesting work for the total four classes related to the 11 closed Arabic characters (out of the 28 characters of the Arabic alphabet)..