Segmentation and restoration of images of old books

Hatem Hamza (1)

1 READ - READ

Today, thousands of old books remain unused all over the world. This is a big loss for humanity and future generations. The state of these books is getting worse everyday because their conservation techniques are often poor. The goal of this project is to try to restore the images of old books. I will give examples from a book from the renaissance on which I have already worked before and other examples fro the Gutenberg digital library. The ultimate goal of their restoration is to be able to build digital libraries. This is very important as it enables future generations to discover the large heritage from previous ones. The problem with images of old books is the poor quality of the books. This is the reason why we always have transparent images of the writing from the back that show on the front and vice versa. This is one of the main problems in images of old books restoration. Many approaches have been proposed on different types of documents. I am going to focus my discussion on the use of Self Organizing maps (SOM) for this project.

Mots clés: classification d'images de documents; document image classification; kohonen map; carte de kohonen; kmeans

Domaines: Autre [cs.OH]

Hatem Hamza. Segmentation and restoration of images of old books. [Intern report] A04-R-063 || hamza04a, 2004, 10 p. ⟨inria-00100217⟩