A hybrid feature dimension reduction approach for image classification

Qi Tian, Jie Yu, Ting Rui, Thomas S Huang

Research output: Contribution to journalConference article

Abstract

In content-based image retrieval (CBIR), in order to alleviate learning in the high-dimensional space, Fisher discriminant analysis (FDA) and multiple discriminant analysis (MDA) are commonly used to find an optimal discriminating subspace that the data are clustered in the reduced feature space, in which the probabilistic structure of the data could be simplified and captured by simpler model assumption, e.g., Gaussian mixtures. However, due to the two reasons (i) the real number of classes in the image database is usually unknown; and (ii) the image retrieval system acts as a classifier to divide the images into two classes, relevant and irrelevant, the effective dimension of projected subspace is usually one. In this paper, a novel hybrid feature dimension reduction technique is proposed to construct descriptive and discriminant features at the same time by maximizing the Rayleigh coefficient. The hybrid LDA and PCA analysis not only increases the effective dimension of the projected subspace, but also offers more flexibility and alternatives to LDA and PCA. Extensive tests on benchmark and real image databases have shown the superior performance of the hybrid analysis.

Original languageEnglish (US)
Article number02
Pages (from-to)13-24
Number of pages12
JournalProceedings of SPIE - The International Society for Optical Engineering
Volume5601
DOIs
StatePublished - Dec 1 2004
EventInternet Multimedia Management Systems V - Philadelphia, PA, United States
Duration: Oct 26 2004Oct 28 2004

Fingerprint

image classification
Image classification
Image Classification
Image retrieval
Discriminant analysis
Dimension Reduction
Effective Dimension
Image Database
Subspace
Fisher Discriminant Analysis
Classifiers
Gaussian Mixture
Content-based Image Retrieval
retrieval
Image Retrieval
Feature Space
Discriminant Analysis
Discriminant
Rayleigh
Divides

Keywords

  • Dimension reduction
  • Hybrid analysis
  • Image classification
  • LDA
  • PCA

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Computer Science Applications
  • Applied Mathematics
  • Electrical and Electronic Engineering

Cite this

A hybrid feature dimension reduction approach for image classification. / Tian, Qi; Yu, Jie; Rui, Ting; Huang, Thomas S.

In: Proceedings of SPIE - The International Society for Optical Engineering, Vol. 5601, 02, 01.12.2004, p. 13-24.

Research output: Contribution to journalConference article

@article{f3e90cc60c3e45b8be9428a53ca1b7ed,
title = "A hybrid feature dimension reduction approach for image classification",
abstract = "In content-based image retrieval (CBIR), in order to alleviate learning in the high-dimensional space, Fisher discriminant analysis (FDA) and multiple discriminant analysis (MDA) are commonly used to find an optimal discriminating subspace that the data are clustered in the reduced feature space, in which the probabilistic structure of the data could be simplified and captured by simpler model assumption, e.g., Gaussian mixtures. However, due to the two reasons (i) the real number of classes in the image database is usually unknown; and (ii) the image retrieval system acts as a classifier to divide the images into two classes, relevant and irrelevant, the effective dimension of projected subspace is usually one. In this paper, a novel hybrid feature dimension reduction technique is proposed to construct descriptive and discriminant features at the same time by maximizing the Rayleigh coefficient. The hybrid LDA and PCA analysis not only increases the effective dimension of the projected subspace, but also offers more flexibility and alternatives to LDA and PCA. Extensive tests on benchmark and real image databases have shown the superior performance of the hybrid analysis.",
keywords = "Dimension reduction, Hybrid analysis, Image classification, LDA, PCA",
author = "Qi Tian and Jie Yu and Ting Rui and Huang, {Thomas S}",
year = "2004",
month = "12",
day = "1",
doi = "10.1117/12.571532",
language = "English (US)",
volume = "5601",
pages = "13--24",
journal = "Proceedings of SPIE - The International Society for Optical Engineering",
issn = "0277-786X",
publisher = "SPIE",

}

TY - JOUR

T1 - A hybrid feature dimension reduction approach for image classification

AU - Tian, Qi

AU - Yu, Jie

AU - Rui, Ting

AU - Huang, Thomas S

PY - 2004/12/1

Y1 - 2004/12/1

N2 - In content-based image retrieval (CBIR), in order to alleviate learning in the high-dimensional space, Fisher discriminant analysis (FDA) and multiple discriminant analysis (MDA) are commonly used to find an optimal discriminating subspace that the data are clustered in the reduced feature space, in which the probabilistic structure of the data could be simplified and captured by simpler model assumption, e.g., Gaussian mixtures. However, due to the two reasons (i) the real number of classes in the image database is usually unknown; and (ii) the image retrieval system acts as a classifier to divide the images into two classes, relevant and irrelevant, the effective dimension of projected subspace is usually one. In this paper, a novel hybrid feature dimension reduction technique is proposed to construct descriptive and discriminant features at the same time by maximizing the Rayleigh coefficient. The hybrid LDA and PCA analysis not only increases the effective dimension of the projected subspace, but also offers more flexibility and alternatives to LDA and PCA. Extensive tests on benchmark and real image databases have shown the superior performance of the hybrid analysis.

AB - In content-based image retrieval (CBIR), in order to alleviate learning in the high-dimensional space, Fisher discriminant analysis (FDA) and multiple discriminant analysis (MDA) are commonly used to find an optimal discriminating subspace that the data are clustered in the reduced feature space, in which the probabilistic structure of the data could be simplified and captured by simpler model assumption, e.g., Gaussian mixtures. However, due to the two reasons (i) the real number of classes in the image database is usually unknown; and (ii) the image retrieval system acts as a classifier to divide the images into two classes, relevant and irrelevant, the effective dimension of projected subspace is usually one. In this paper, a novel hybrid feature dimension reduction technique is proposed to construct descriptive and discriminant features at the same time by maximizing the Rayleigh coefficient. The hybrid LDA and PCA analysis not only increases the effective dimension of the projected subspace, but also offers more flexibility and alternatives to LDA and PCA. Extensive tests on benchmark and real image databases have shown the superior performance of the hybrid analysis.

KW - Dimension reduction

KW - Hybrid analysis

KW - Image classification

KW - LDA

KW - PCA

UR - http://www.scopus.com/inward/record.url?scp=17644362677&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=17644362677&partnerID=8YFLogxK

U2 - 10.1117/12.571532

DO - 10.1117/12.571532

M3 - Conference article

AN - SCOPUS:17644362677

VL - 5601

SP - 13

EP - 24

JO - Proceedings of SPIE - The International Society for Optical Engineering

JF - Proceedings of SPIE - The International Society for Optical Engineering

SN - 0277-786X

M1 - 02

ER -