A fast iterative kernel PCA feature extraction for hyperspectral images

Wenzhi Liao, Aleksandra Pizurica, Wilfried Philips, Youguo Pi, Bonnie Law (Editor)

Research output: Contribution to conferencePaper

12 Citations (Scopus)

Abstract

A fast iterative Kernel Principal Component Analysis (KPCA) is proposed to extract features from hyperspectral images. The proposed method is a kernel version of the Candid Covariance-Free Incremental Principal Component Analysis, which solves the eigenvectors through iteration. Without performing eigen decomposition on Gram matrix, our method can reduce the space complexity and time complexity greatly. Experimental results were validated in comparison with the standard KPCA and linear version methods.

Conference

Conference2010 IEEE 17th International conference on Image Processing (ICIP 2010)
Abbreviated titleICIP 2010
CountryChina
CityHong Kong
Period26/09/1030/09/10

Fingerprint

principal components analysis
pattern recognition
principal component analysis
matrix methods
iteration
eigenvectors
decomposition
matrix
method

Keywords

  • kernel version
  • hyperspectral images
  • feature extraction
  • incremental principal component analysis
  • complexity theory
  • computational complexity
  • eigenvalues and eigenfunctions
  • geophysical image processing

Cite this

Liao, W., Pizurica, A., Philips, W., Pi, Y., & Law, B. (Ed.) (2010). A fast iterative kernel PCA feature extraction for hyperspectral images. 1317-1320. Paper presented at 2010 IEEE 17th International conference on Image Processing (ICIP 2010), Hong Kong, China. https://doi.org/10.1109/ICIP.2010.5651670
Liao, Wenzhi ; Pizurica, Aleksandra ; Philips, Wilfried ; Pi, Youguo ; Law, Bonnie (Editor). / A fast iterative kernel PCA feature extraction for hyperspectral images. Paper presented at 2010 IEEE 17th International conference on Image Processing (ICIP 2010), Hong Kong, China.4 p.
@conference{dc7e4b4ef30a456590a6cef752d0caae,
title = "A fast iterative kernel PCA feature extraction for hyperspectral images",
abstract = "A fast iterative Kernel Principal Component Analysis (KPCA) is proposed to extract features from hyperspectral images. The proposed method is a kernel version of the Candid Covariance-Free Incremental Principal Component Analysis, which solves the eigenvectors through iteration. Without performing eigen decomposition on Gram matrix, our method can reduce the space complexity and time complexity greatly. Experimental results were validated in comparison with the standard KPCA and linear version methods.",
keywords = "kernel version, hyperspectral images, feature extraction, incremental principal component analysis, complexity theory, computational complexity, eigenvalues and eigenfunctions, geophysical image processing",
author = "Wenzhi Liao and Aleksandra Pizurica and Wilfried Philips and Youguo Pi and Bonnie Law",
year = "2010",
month = "12",
day = "3",
doi = "10.1109/ICIP.2010.5651670",
language = "English",
pages = "1317--1320",
note = "2010 IEEE 17th International conference on Image Processing (ICIP 2010), ICIP 2010 ; Conference date: 26-09-2010 Through 30-09-2010",

}

Liao, W, Pizurica, A, Philips, W, Pi, Y & Law, B (ed.) 2010, 'A fast iterative kernel PCA feature extraction for hyperspectral images' Paper presented at 2010 IEEE 17th International conference on Image Processing (ICIP 2010), Hong Kong, China, 26/09/10 - 30/09/10, pp. 1317-1320. https://doi.org/10.1109/ICIP.2010.5651670

A fast iterative kernel PCA feature extraction for hyperspectral images. / Liao, Wenzhi; Pizurica, Aleksandra; Philips, Wilfried; Pi, Youguo; Law, Bonnie (Editor).

2010. 1317-1320 Paper presented at 2010 IEEE 17th International conference on Image Processing (ICIP 2010), Hong Kong, China.

Research output: Contribution to conferencePaper

TY - CONF

T1 - A fast iterative kernel PCA feature extraction for hyperspectral images

AU - Liao, Wenzhi

AU - Pizurica, Aleksandra

AU - Philips, Wilfried

AU - Pi, Youguo

A2 - Law, Bonnie

PY - 2010/12/3

Y1 - 2010/12/3

N2 - A fast iterative Kernel Principal Component Analysis (KPCA) is proposed to extract features from hyperspectral images. The proposed method is a kernel version of the Candid Covariance-Free Incremental Principal Component Analysis, which solves the eigenvectors through iteration. Without performing eigen decomposition on Gram matrix, our method can reduce the space complexity and time complexity greatly. Experimental results were validated in comparison with the standard KPCA and linear version methods.

AB - A fast iterative Kernel Principal Component Analysis (KPCA) is proposed to extract features from hyperspectral images. The proposed method is a kernel version of the Candid Covariance-Free Incremental Principal Component Analysis, which solves the eigenvectors through iteration. Without performing eigen decomposition on Gram matrix, our method can reduce the space complexity and time complexity greatly. Experimental results were validated in comparison with the standard KPCA and linear version methods.

KW - kernel version

KW - hyperspectral images

KW - feature extraction

KW - incremental principal component analysis

KW - complexity theory

KW - computational complexity

KW - eigenvalues and eigenfunctions

KW - geophysical image processing

UR - http://hdl.handle.net/1854/LU-1058464

U2 - 10.1109/ICIP.2010.5651670

DO - 10.1109/ICIP.2010.5651670

M3 - Paper

SP - 1317

EP - 1320

ER -

Liao W, Pizurica A, Philips W, Pi Y, Law B, (ed.). A fast iterative kernel PCA feature extraction for hyperspectral images. 2010. Paper presented at 2010 IEEE 17th International conference on Image Processing (ICIP 2010), Hong Kong, China. https://doi.org/10.1109/ICIP.2010.5651670