الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 228 |  |  |
| | | from | 228 | | |
Abstract
The binarization process of historical manuscripts images is one of the most important processes because there are different applications were based on its quality, as Optical character Recognition (OCR), and other important applications. This process was selected in this thesis, which the historical manuscripts contain important and valuable information in the various fields of knowledge sciences and they are considered a true treasure in the transmission of science and knowledge from generation to generation. There are some problems associated with these manuscripts which arise from certain factors as poor preservation and weather factors, etc..., which leads to the emergence of some problems (e.g. variable background intensity, bleed-through, shadows, low contrast, smudge, and smear). where an input image of a degraded document is converted to a binary image for separating the text from the background regions. Thus, in this thesis, this topic was chosen because it represents great importance for the preservation of heritage. We mean that the treatment of historical manuscripts will be more useful and easier for using and allowing every user to access all information easily and accurately than manual use (traditional methods). Images of historical Arabic manuscripts were collected and ground truth images are created. In addition, images from the competition (H) -DIBCO (2009-2016) were also used. H-DIBCO evaluation measures were used to assess the efficiency of the proposed approaches, including a visual inspection. We depend on the optimization algorithms in this thesis. There are different proposed approaches designed for addressing the binarization problem which was treated as an optimization problem. Four different approaches were proposed which based on different optimization algorithms, which the k-means objective function was used. Furthermore, ten chaotic maps were used to adjust the parameters on the basic version of the Ant-lion optimization algorithm (ALO) and Salp Swarm Algorithm (SSA). Many experiments have been conducted to evaluate the output from the proposed approaches. The evaluation process was done based on the performance measures values, visual inspection, and the convergence curves of the optimization algorithms. The experimental results indicated the superiority of the proposed approaches which are able to deal with different types of noises efficiently, that produces high-quality binarized images.