Unsupervised learning of a finite mixture model based on the Dirichlet distribution and its application

• Published in: IEEE transactions on image processing Vol. 13; no. 11; pp. 1533 - 1543
• Main Authors: Bouguila, N., Ziou, D., Vaillancourt, J.
• Format: Journal Article
• Language: English
• Published: New York, NY IEEE 01.11.2004; Institute of Electrical and Electronics Engineers; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Applied sciences; Artificial Intelligence; Cluster Analysis; Colorimetry - methods; Computer Graphics; Computer Simulation; Detection, estimation, filtering, equalization, prediction; Dirichlet problem; Exact sciences and technology; Flexibility; Histograms; Humans; Image databases; Image Enhancement - methods; Image Interpretation, Computer-Assisted - methods; Image retrieval; Imaging, Three-Dimensional - methods; Information retrieval; Information Storage and Retrieval - methods; Information, signal and communications theory; Learning; Mathematical analysis; Mathematical models; Maximum likelihood detection; Maximum likelihood estimation; Models, Biological; Models, Statistical; Numerical Analysis, Computer-Assisted; Parameter estimation; Pattern Recognition, Automated - methods; Reproducibility of Results; Scoring; Sensitivity and Specificity; Signal and communications theory; Signal Processing, Computer-Assisted; Signal representation. Spectral analysis; Signal, noise; Skin; Skin Pigmentation - physiology; Statistical Distributions; Subtraction Technique; Telecommunications and information theory; Unsupervised learning; User-Computer Interface; Riemannian space; Multivariate process; Parameter estimation; Histogram; Image databank; Multimedia databases; Mixture theory; Abstracting; Unsupervised classification; Signal classification; Flexibility; Modeling; Dirichlet distribution; Learning; mixture modeling; natural gradient; Boundary value problem; Dirichlet problem; Fisher's scoring method; image summarizing; Maximum likelihood; Learning algorithm
• ISSN: 1057-7149; 1941-0042
• DOI: 10.1109/TIP.2004.834664

This paper presents an unsupervised algorithm for learning a finite mixture model from multivariate data. This mixture model is based on the Dirichlet distribution, which offers high flexibility for modeling data. The proposed approach for estimating the parameters of a Dirichlet mixture is based on the maximum likelihood (ML) and Fisher scoring methods. Experimental results are presented for the following applications: estimation of artificial histograms, summarization of image databases for efficient retrieval, and human skin color modeling and its application to skin detection in multimedia databases.