|
|||
|
A New Unsupervised Color Image Segmentation Algorithm upon
a Statistical Multidimensional Data Analysis Approach
A. Hamid (1) , R. Allaoui (2) , &
A. Sbihi (2)
(1) University
Mohammed V Agdal, Faculté des Sciences, LETS,
BP 1014, Rabat 1000, Morocco.
(2) University
Ibn Tofaïl, Faculté des Sciences, LIRF,
BP 133, Kénitra 14000, Morocco.
Abstract:
Cluster analysis is an important technique in exploratory
data analysis when no a priori knowledge of the distribution
of the observed data is available. Clustering methods,
that divide the data into natural classes, have been
used in a large variety of engineering and scientific
domains such as pattern recognition, learning theory,
astrophysics, and image processing. A new alternative
for unsupervised color image segmentation is given as
application of a new cluster analysis technique, proposed
in this work, dealing with mode detection of the underlying
probability density function (pdf) . A recursive separable
hyperbolic filter, which is with a reliability criterion
allowing modeling as well the pdf variations as the
noise attached to it, is generalized to a multidimensional
space. Based on the characteristic theorem of convexity,
the proposed technique leads to extract automatically
modal regions as concave connected components corresponding
to the clusters in the mixture. The observations falling
in the so-detected modal regions are taken as prototypes
for classification.
Key words:
Hyperbolic operator,
mode detection, cluster analysis, color image segmentation.
Biography:
Ahmed Hamid is a telecommunications engineer since 1981. He received, in 1988, the Diploma of the Superior Studies in Transmission and Signal Processing from the University Mohammed V of Rabat, Morocco. Since 1999, his research interests have been oriented toward the pattern classification and image processing. He is now, in parallel to his engineer's function, a researcher at the Electronic and Signal Processing Laboratory in the University Mohammed V of Rabat. Rabha Allaoui received the Licence of Science degree in Applied Mathematics from the University Quadi Ayyad of Marrakech, Morocco, in 1995, and the PhD in Automatic Control and Information Theory from the University Ibn Tofaïl of Kénitra, Morocco, in 2002. She worked two years as a researcher at the Moroccan Ministry of the public health in Rabat. She is now an Assistant Professor at the University Hassan Premier, Settat, Morocco. Her current research interests are in the multidimensional data analysis and image processing. Abderrahmane Sbihi received the "Licence" of Science degree in physics from the University Mohammed V of Rabat, Morocco, in 1983 and the "Docteur ès Sciences" degree from the university Ibn Tofaïl of Kénitra, Morocco, in 1995. In 1985, he joined this last university where he worked as an Assistant Professor. From 1992 to 1995, he joined the University of Science and Technology of Lille, France, as a Guest Assistant Professor at the Institute of Technology and as a Research Assistant in the Automatic Control Center where he worked on the adaptation of numerical mathematical morphology to pattern classification problems. In 1995, he returned to the university Ibn Tofaïl where he is currently Professor and Head of the Computer Vision and Pattern Recognition Laboratory (LIRF). His current research interests include computer vision, pattern recognition and multidimensional data analysis. BibTex: @ARTICLE{P1150510002,
AUTHOR = {A. Hamid and R. Allaoui and and A. Sbihi},
TITLE = {A New Unsupervised Color Image Segmentation Algorithm upon a Statistical Multidimensional Data Analysis Approach}, JOURNAL = {ICGST International Journal on Graphics, Vision and Image Processing}, YEAR = {2005}, MONTH = {April}, VOLUME = {05}, ISSUE = {4}, PAGES = {33--40} } (
|
|||
|
|
