A new kernel method for hyperspectral image feature extraction

Bin Zhao, Lianru Gao, Wenzhi Liao, Bing Zhang, Xin Huang, Jiayi Li (Editor), Jocelyn Chanussot

Research output: Contribution to journalArticle

18 Citations (Scopus)
7 Downloads (Pure)

Abstract

Hyperspectral image provides abundant spectral information for remote discrimination of subtle differences in ground covers. However, the increasing spectral dimensions, as well as the information redundancy, make the analysis and interpretation of hyperspectral images a challenge. Feature extraction is a very important step for hyperspectral image processing. Feature extraction methods aim at reducing the dimension of data, while preserving as much information as possible. Particularly, nonlinear feature extraction methods (e.g. kernel minimum noise fraction (KMNF) transformation) have been reported to benefit many applications of hyperspectral remote sensing, due to their good preservation of high-order structures of the original data. However, conventional KMNF or its extensions have some limitations on noise fraction estimation during the feature extraction, and this leads to poor performances for postapplications. This paper proposes a novel nonlinear feature extraction method for hyperspectral images. Instead of estimating noise fraction by the nearest neighborhood information (within a sliding window), the proposed method explores the use of image segmentation. The approach benefits both noise fraction estimation and information preservation, and enables a significant improvement for classification. Experimental results on two real hyperspectral images demonstrate the efficiency of the proposed method. Compared to conventional KMNF, the improvements of the method on two hyperspectral image classification are 8 and 11%. This nonlinear feature extraction method can be also applied to other disciplines where highdimensional data analysis is required.
Original languageEnglish
Pages (from-to)309-318
Number of pages10
JournalGeo-spatial Information Science
Volume20
Issue number4
DOIs
Publication statusPublished - 4 Dec 2017

Fingerprint

pattern recognition
extraction method
image classification
ground cover
redundancy
image processing
segmentation
preserving
sliding
discrimination
method
remote sensing
estimating

Keywords

  • principal companion
  • noise estimation
  • classification
  • algorithm
  • selection
  • band
  • hyperspectral image

Cite this

Zhao, B., Gao, L., Liao, W., Zhang, B., Huang, X., Li, J. (Ed.), & Chanussot, J. (2017). A new kernel method for hyperspectral image feature extraction. Geo-spatial Information Science, 20(4), 309-318. https://doi.org/10.1080/10095020.2017.1403088
Zhao, Bin ; Gao, Lianru ; Liao, Wenzhi ; Zhang, Bing ; Huang, Xin ; Li, Jiayi (Editor) ; Chanussot, Jocelyn. / A new kernel method for hyperspectral image feature extraction. In: Geo-spatial Information Science. 2017 ; Vol. 20, No. 4. pp. 309-318.
@article{f49bd1d7cc6143838b4cf494e397d8cd,
title = "A new kernel method for hyperspectral image feature extraction",
abstract = "Hyperspectral image provides abundant spectral information for remote discrimination of subtle differences in ground covers. However, the increasing spectral dimensions, as well as the information redundancy, make the analysis and interpretation of hyperspectral images a challenge. Feature extraction is a very important step for hyperspectral image processing. Feature extraction methods aim at reducing the dimension of data, while preserving as much information as possible. Particularly, nonlinear feature extraction methods (e.g. kernel minimum noise fraction (KMNF) transformation) have been reported to benefit many applications of hyperspectral remote sensing, due to their good preservation of high-order structures of the original data. However, conventional KMNF or its extensions have some limitations on noise fraction estimation during the feature extraction, and this leads to poor performances for postapplications. This paper proposes a novel nonlinear feature extraction method for hyperspectral images. Instead of estimating noise fraction by the nearest neighborhood information (within a sliding window), the proposed method explores the use of image segmentation. The approach benefits both noise fraction estimation and information preservation, and enables a significant improvement for classification. Experimental results on two real hyperspectral images demonstrate the efficiency of the proposed method. Compared to conventional KMNF, the improvements of the method on two hyperspectral image classification are 8 and 11{\%}. This nonlinear feature extraction method can be also applied to other disciplines where highdimensional data analysis is required.",
keywords = "principal companion, noise estimation, classification, algorithm, selection, band, hyperspectral image",
author = "Bin Zhao and Lianru Gao and Wenzhi Liao and Bing Zhang and Xin Huang and Jiayi Li and Jocelyn Chanussot",
year = "2017",
month = "12",
day = "4",
doi = "10.1080/10095020.2017.1403088",
language = "English",
volume = "20",
pages = "309--318",
journal = "Geo-spatial Information Science",
issn = "1009-5020",
number = "4",

}

Zhao, B, Gao, L, Liao, W, Zhang, B, Huang, X, Li, J (ed.) & Chanussot, J 2017, 'A new kernel method for hyperspectral image feature extraction', Geo-spatial Information Science, vol. 20, no. 4, pp. 309-318. https://doi.org/10.1080/10095020.2017.1403088

A new kernel method for hyperspectral image feature extraction. / Zhao, Bin; Gao, Lianru; Liao, Wenzhi; Zhang, Bing; Huang, Xin; Li, Jiayi (Editor); Chanussot, Jocelyn.

In: Geo-spatial Information Science, Vol. 20, No. 4, 04.12.2017, p. 309-318.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A new kernel method for hyperspectral image feature extraction

AU - Zhao, Bin

AU - Gao, Lianru

AU - Liao, Wenzhi

AU - Zhang, Bing

AU - Huang, Xin

AU - Chanussot, Jocelyn

A2 - Li, Jiayi

PY - 2017/12/4

Y1 - 2017/12/4

N2 - Hyperspectral image provides abundant spectral information for remote discrimination of subtle differences in ground covers. However, the increasing spectral dimensions, as well as the information redundancy, make the analysis and interpretation of hyperspectral images a challenge. Feature extraction is a very important step for hyperspectral image processing. Feature extraction methods aim at reducing the dimension of data, while preserving as much information as possible. Particularly, nonlinear feature extraction methods (e.g. kernel minimum noise fraction (KMNF) transformation) have been reported to benefit many applications of hyperspectral remote sensing, due to their good preservation of high-order structures of the original data. However, conventional KMNF or its extensions have some limitations on noise fraction estimation during the feature extraction, and this leads to poor performances for postapplications. This paper proposes a novel nonlinear feature extraction method for hyperspectral images. Instead of estimating noise fraction by the nearest neighborhood information (within a sliding window), the proposed method explores the use of image segmentation. The approach benefits both noise fraction estimation and information preservation, and enables a significant improvement for classification. Experimental results on two real hyperspectral images demonstrate the efficiency of the proposed method. Compared to conventional KMNF, the improvements of the method on two hyperspectral image classification are 8 and 11%. This nonlinear feature extraction method can be also applied to other disciplines where highdimensional data analysis is required.

AB - Hyperspectral image provides abundant spectral information for remote discrimination of subtle differences in ground covers. However, the increasing spectral dimensions, as well as the information redundancy, make the analysis and interpretation of hyperspectral images a challenge. Feature extraction is a very important step for hyperspectral image processing. Feature extraction methods aim at reducing the dimension of data, while preserving as much information as possible. Particularly, nonlinear feature extraction methods (e.g. kernel minimum noise fraction (KMNF) transformation) have been reported to benefit many applications of hyperspectral remote sensing, due to their good preservation of high-order structures of the original data. However, conventional KMNF or its extensions have some limitations on noise fraction estimation during the feature extraction, and this leads to poor performances for postapplications. This paper proposes a novel nonlinear feature extraction method for hyperspectral images. Instead of estimating noise fraction by the nearest neighborhood information (within a sliding window), the proposed method explores the use of image segmentation. The approach benefits both noise fraction estimation and information preservation, and enables a significant improvement for classification. Experimental results on two real hyperspectral images demonstrate the efficiency of the proposed method. Compared to conventional KMNF, the improvements of the method on two hyperspectral image classification are 8 and 11%. This nonlinear feature extraction method can be also applied to other disciplines where highdimensional data analysis is required.

KW - principal companion

KW - noise estimation

KW - classification

KW - algorithm

KW - selection

KW - band

KW - hyperspectral image

U2 - 10.1080/10095020.2017.1403088

DO - 10.1080/10095020.2017.1403088

M3 - Article

VL - 20

SP - 309

EP - 318

JO - Geo-spatial Information Science

JF - Geo-spatial Information Science

SN - 1009-5020

IS - 4

ER -