Towards Better Resolution of Ottoman Archive Documents

It is very important to transfer the enormous archive left by the Ottomans to the present and the future. Huge digital archive transformation and collection studies are carried out in this context. Resolution of digitized archive documents can affect the performance and quality of the studies in text search, text analysis, character recognition, etc. In this study, a deep learning method tried to improve the resolution of historical text images and thus the image quality. The algorithm was trained to enlarge images for the scale factors 2, 3 and 4. The performance of the algorithm was measured quantitatively, and qualitatively in terms of the human visual system. The proposed method obtained very cheering results in terms of image quality and visual quality. The qualitative and quantitative evaluations have shown that the resolution of historical text images can be increased very successfully with deep learning approach. Furthermore, such algorithms can assist also in more successful processing and analysis of historical Ottoman documents.