THE FIRST INTERNATIONAL NEWSPAPER SEGMENTATION CONTEST
at ICDAR'2001
http://www.lpa.gr/contest
CALL FOR PARTICIPATION
The aim of this contest is to evaluate all existing
algorithms for document image segmentation that can be
applied to Newspaper page segmentation. Many algorithms
for page segmentation have been proposed by several
research groups dealing with document image processing.
However, many of these algorithms are not directly
applicable to newspaper images, which present special
problems. The most significant problems identified
include the complex layout of newspaper pages,
particularly the oldest ones, where text columns are
located very close to each other in a haphazard way, as
well as the poor scanning results derived from paper
material that was originally of low print quality or has
deteriorated through time. Another important problem for
newspaper page segmentation is the layout habits that
seem to have changed repeatedly.
We have selected to evaluate the tracing of all basic
entities that appear in most newspaper pages and cover a
time period from the beginning of the previous century
till today. We selected entities that belong to text
regions, lines and images/drawings.
Training and test sets:
Both training and test sets come from Greek and English
newspapers. The Greek newspapers are "TO VIMA" (THE
TRIBUNE) and "ELETHERON VIMA" (FREE TRIBUNE) and we have
selected front pages from years 1922, 1940 and 1968. Both
newspapers are published by Lambrakis Press S.A.. The
English speaking newspaper is the "International Herald
Tribune" and we have selected front pages from years
1900, 1925 and 1950.
Performance Evaluation:
The performance evaluation method that will be used is
based on counting the number of matches between the
entities detected by the algorithm and the entities in
the ground truth. We use a global MatchScore table for
all entities whose values are calculated according to the
intersection of the ON pixel sets of the result and the
ground truth. In order to evaluate the detection of
newspaper page components, we employ a global performance
metric that combines the average values of detection rate
and recognition accuracy.
Participating benefits:
We will collect all segmentation results and
automatically rank all efforts. The best working
algorithms will appear in a report that will be prepared
by the contest organisers. Authors of the highest ranking
algorithms will be invited to prepare papers that will be
included at ICDAR 2001 proceedings and presented in an
oral session during the conference.
Important dates:
Register Interest & Download Training set: ASAP
Download Test set: May 1, 2001
E-mail results and algorithm description: May 14,2001
Results announcement & notification to participants
with best working algorithms to prepare a paper: May 16,2001
Submission of all papers: June 4, 2001
Notification of acceptance: June 11, 2001
Camera-ready manuscripts received: June 30, 2001
Organizing Committee:
Dr Basilis Gatos
Head of Research
Lambrakis Press Archives
Department of Digital Technologies
8, Heyden Street,
104 34 Athens, Greece
Tel: +30-1-8252680, +30-1-8252540
Fax: +30-1-8250040
E-mail: bgatos@lpa.gr
URL: http://www.lpa.gr
Dr Apostolos Antonacopoulos
Lecturer, Dept. of Computer Science
The University of Liverpool
Chadwick Building, Peach Street
Liverpool, L69 7ZF, U.K.
Tel.: +44 (151) 794 3695
Fax.: +44 (151) 794 3715
E-mail: aa@csc.liv.ac.uk
URL: http://www.csc.liv.ac.uk/~aa