Université de Lyon (92 rue Pasteur - CS 30122, 69361 Lyon Cedex 07 - France)
Abstract : Historical documents are valuable resources worth
to be preserved in order to support our cultural and social knowledge. Unfortunately, these supports based on fragile materials are often affected by several types of degradations. Applying restoration techniques on
degraded captured digital images of historical documents may be a quick and efficient way to preserve the document and avoid the loss in its content.
This paper presents a new method to restore a
particular type of degradation which is referred to as
"bleed-through". This degradation is caused by the interference of characters from the reverse side with
the text to be read. Our proposed method is based on a recursive approach that relies on two types of analysis:
the Principal Component Analysis and the k-means
clustering algorithm. The aim here is to extract clear
textural images from these interfering and overlapping areas of text. Our restoration method analyses the
front side image alone and corrects the unneeded
image components. This paper concludes with some
experimental results that demonstrate the effectiveness of our proposed method.
https://hal.archives-ouvertes.fr/hal-01505740
Contributor : Équipe Gestionnaire Des Publications Si Liris <>
Submitted on : Tuesday, April 11, 2017 - 4:56:29 PM Last modification on : Wednesday, July 8, 2020 - 12:42:08 PM
Fadoua Drira, Hubert Emptoz. A Recursive Approach For Bleed-Through removal. First International Workshop on Camera-Based Document Analysis and Recognition (CBDAR2005), Aug 2005, Seoul, South Korea. pp.119-126. ⟨hal-01505740⟩