Fast Haar transform based feature extraction for face representation and recognition

Yanwei Pang*, Xuelong Li, Yuan Yuan, Dacheng Tao, Jing Pan

*Corresponding author for this work

Research output: Contribution to journalArticle

Abstract

Subspace learning is the process of finding a proper feature subspace and then projecting high-dimensional data onto the learned low-dimensional subspace. The projection operation requires many floating-point multiplications and additions, which makes the projection process computationally expensive. To tackle this problem, this paper proposes two simple-but-effective fast subspace learning and image projection methods, fast Haar transform (FHT) based principal component analysis and FHT based spectral regression discriminant analysis. The advantages of these two methods result from employing both the FHT for subspace learning and the integral vector for feature extraction. Experimental results on three face databases demonstrated their effectiveness and efficiency.

Original languageEnglish
Pages (from-to)441-450
Number of pages10
JournalIEEE Transactions on Information Forensics and Security
Volume4
Issue number3
Early online date7 Jul 2009
DOIs
Publication statusPublished - Sep 2009

Fingerprint

Discriminant analysis
Principal component analysis
Feature extraction

Keywords

  • face representation and recognition
  • fast algorithm
  • feature extraction
  • Haar transform
  • subspace analysis

Cite this

Pang, Yanwei ; Li, Xuelong ; Yuan, Yuan ; Tao, Dacheng ; Pan, Jing. / Fast Haar transform based feature extraction for face representation and recognition. In: IEEE Transactions on Information Forensics and Security. 2009 ; Vol. 4, No. 3. pp. 441-450.
@article{dde8997d6a294e76aa61e2c84a6871bf,
title = "Fast Haar transform based feature extraction for face representation and recognition",
abstract = "Subspace learning is the process of finding a proper feature subspace and then projecting high-dimensional data onto the learned low-dimensional subspace. The projection operation requires many floating-point multiplications and additions, which makes the projection process computationally expensive. To tackle this problem, this paper proposes two simple-but-effective fast subspace learning and image projection methods, fast Haar transform (FHT) based principal component analysis and FHT based spectral regression discriminant analysis. The advantages of these two methods result from employing both the FHT for subspace learning and the integral vector for feature extraction. Experimental results on three face databases demonstrated their effectiveness and efficiency.",
keywords = "face representation and recognition, fast algorithm, feature extraction, Haar transform, subspace analysis",
author = "Yanwei Pang and Xuelong Li and Yuan Yuan and Dacheng Tao and Jing Pan",
year = "2009",
month = "9",
doi = "10.1109/TIFS.2009.2026455",
language = "English",
volume = "4",
pages = "441--450",
journal = "IEEE Transactions on Information Forensics and Security",
issn = "1556-6013",
publisher = "IEEE",
number = "3",

}

Fast Haar transform based feature extraction for face representation and recognition. / Pang, Yanwei; Li, Xuelong; Yuan, Yuan; Tao, Dacheng; Pan, Jing.

In: IEEE Transactions on Information Forensics and Security, Vol. 4, No. 3, 09.2009, p. 441-450.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Fast Haar transform based feature extraction for face representation and recognition

AU - Pang, Yanwei

AU - Li, Xuelong

AU - Yuan, Yuan

AU - Tao, Dacheng

AU - Pan, Jing

PY - 2009/9

Y1 - 2009/9

N2 - Subspace learning is the process of finding a proper feature subspace and then projecting high-dimensional data onto the learned low-dimensional subspace. The projection operation requires many floating-point multiplications and additions, which makes the projection process computationally expensive. To tackle this problem, this paper proposes two simple-but-effective fast subspace learning and image projection methods, fast Haar transform (FHT) based principal component analysis and FHT based spectral regression discriminant analysis. The advantages of these two methods result from employing both the FHT for subspace learning and the integral vector for feature extraction. Experimental results on three face databases demonstrated their effectiveness and efficiency.

AB - Subspace learning is the process of finding a proper feature subspace and then projecting high-dimensional data onto the learned low-dimensional subspace. The projection operation requires many floating-point multiplications and additions, which makes the projection process computationally expensive. To tackle this problem, this paper proposes two simple-but-effective fast subspace learning and image projection methods, fast Haar transform (FHT) based principal component analysis and FHT based spectral regression discriminant analysis. The advantages of these two methods result from employing both the FHT for subspace learning and the integral vector for feature extraction. Experimental results on three face databases demonstrated their effectiveness and efficiency.

KW - face representation and recognition

KW - fast algorithm

KW - feature extraction

KW - Haar transform

KW - subspace analysis

UR - http://ieeexplore.ieee.org/document/5159453/

UR - http://www.scopus.com/inward/record.url?scp=69749118031&partnerID=8YFLogxK

U2 - 10.1109/TIFS.2009.2026455

DO - 10.1109/TIFS.2009.2026455

M3 - Article

AN - SCOPUS:69749118031

VL - 4

SP - 441

EP - 450

JO - IEEE Transactions on Information Forensics and Security

JF - IEEE Transactions on Information Forensics and Security

SN - 1556-6013

IS - 3

ER -