By Topic

The Unsupervised Niche Clustering algorithm: extension to multivariate clusters and application to color image segmentation

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
O. Nasraoui ; Dept. of Electr. & Comput. Eng., Memphis State Univ., TN, USA ; R. Krishnapuram

The Unsupervised Niche Clustering (UNC) is a new robust clustering algorithm that can successfully find dense areas (clusters) in feature space and determine the number of clusters automatically. The clustering problem is converted to a multimodal function optimization problem within the context of Genetic Niching. However, UNC was only formulated for using the Euclidean distance and for 2-dimensional data sets, which means that it could only be tested on spherically distributed 2-D data. In this paper, we extend UNC to the case of general multivariate distributions in n-D space by modifying our definition of scale and adapting the mating restriction rules and extraction procedure accordingly. Genetic Optimization makes our approach less prone to suboptimal solutions and the use of robust weights makes it less sensitive to the presence of noise. Most importantly, Genetic optimization frees our approach from the necessity of deriving prototype equations when such derivations are impossible, such as when a nondifferentiable distance measure is used. This enables our approach to handle data with both numeric and qualitative attributes, and general subjective non-metric dissimilarity measures. The effectiveness of the extended UNC is demonstrated using several examples to illustrate its performance for data sets with clusters; of various size, density, orientation, and noise contamination rates. The results are also compared with the K-Means and the Possibilistic C-Means

Published in:

IFSA World Congress and 20th NAFIPS International Conference, 2001. Joint 9th  (Volume:3 )

Date of Conference:

25-28 July 2001