Arabic calligraphy recognition based on binarization methods and degraded images

Optical Font Recognition is one of the main challenges in this time. The available methods of optical font recognition are deal with the recent documents and fonts types. However, there are neglected in dealing with the historical and regarded documents. Moreover, they have neglected languages that are not belong into Asian or Latin. Regarding to those types of documents, we proposed a new framework of optical font recognition for Arabic calligraphy. We enhance binarization method based on previous works. By introducing that, we achieve better quality images at the preprocessing stage. Then we generate text block before passing mailing to post-processing stages. Then, we extract the features based on edge direction matrixes. In the classification stage, we apply backpropagation neural network to identify the font type of the calligraphy. We observe that our proposal method achieve better performance in both preprocessing and post processing.