Abstract:
While historians face various challenges in the process of dating the texts of historical manuscripts, computer scientists face multiple difficulties in automating these texts. To address this problem, deep learning techniques that have proven effectiveness in other fields have been used. Of this study presents the various pre-processing methods used in character recognition systems, which cater to a wide range of image types, as these images include simple handwritten forms and documents with colorful and complex backgrounds and varying intensity. Basic pre-processing techniques are comprehensively discussed, including aberration detection and correction, contrast stretching for image optimization, binary encoding, noise removal methods, normalization, segmentation, and morphological processing techniques
