A New and Efficient Algorithm to Binarize Document Images Removing Back-to-Front Interference

João Marcelo Monte da Silva (Universidade Federal de Pernambuco, Brazil)

Rafael Dueire Lins (Universidade Federal de Pernambuco, Brazil)

Fernando Mário Junqueira Martins (Universidade do Minho, Portugal)

Rosita Wachenchauzer (Universidad de Buenos Aires, Argentina)

Abstract: "Back-to-front interference", "bleeding" and "show-through"is the name given to the phenomenon found whenever documents are written on both sides of translucent paper and the print of one side is visible on the other one. The binarization of documents with back-to-front interference with standard algorithms yields unreadable documents. This paper presents a fast entropy-based segmentation method for generating high-quality binarized images of documents with back-to-front interference.

Keywords: back-to-front interference, bleeding, document engineering, show through

Categories: H.3.3