Abstract
This paper is in continuation of the on-going research project, which clearly focuses on the methodology used to accomplish the research work entitled "Extracting Graphical Information from Arabic e-Documents for Visually Impaired People". The accuracy of the extracted and recognized text is improved using the proposed hybrid traversing text extraction technique. The experiments were conducted on images containing Arabic text and the proposed technique takes less computation time when compared to methods used by the researcher in their earlier works. The results of the research are very promising with a recognition rate of 91.3%.