A parallel unmixing algorithm for hyperspectral images

Stefan Robila, Lukasz G. Maciak

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

12 Citations (Scopus)

Abstract

We present a new algorithm for feature extraction in hyperspectral images based on source separation and parallel computing. In source separation, given a linear mixture of sources, the goal is to recover the components by producing an unmixing matrix. In hyperspectral imagery, the mixing transform and the separated components can be associated with endmembers and their abundances. Source separation based methods have been employed for target detection and classification of hyperspectral images. However, these methods usually involve restrictive conditions on the nature of the results such as orthogonality (in Principal Component Analysis - PCA and Orthogonal Subspace Projection - OSP) of the endmembers or statistical independence (in Independent Component Analysis - ICA) of the abundances nor do they fully satisfy all the conditions included in the Linear Mixing Model. Compared to this, our approach is based on the Nonnegative Matrix Factorization (NMF), a less constraining unmixing method. NMF has the advantage of producing positively defined data, and, with several modifications that we introduce also ensures addition to one. The endmember vectors and the abundances are obtained through a gradient based optimization approach. The algorithm is further modified to run in a parallel environment. The parallel NMF (P-NMF) significantly reduces the time complexity and is shown to also easily port to a distributed environment. Experiments with in-house and Hydice data suggest that NMF outperforms ICA, PCA and OSP for unsupervised endmember extraction. Coupled with its parallel implementation, the new method provides an efficient way for unsupervised unmixing further supporting our efforts in the development of a real time hyperspectral sensing environment with applications to industry and life sciences.

Original languageEnglish
Title of host publicationIntelligent Robots and Computer Vision XXIV
Subtitle of host publicationAlgorithms, Techniques, and Active Vision
Volume6384
DOIs
StatePublished - 28 Nov 2006
EventIntelligent Robots and Computer Vision XXIV: Algorithms, Techniques, and Active Vision - Boston, MA, United States
Duration: 2 Oct 20064 Oct 2006

Other

OtherIntelligent Robots and Computer Vision XXIV: Algorithms, Techniques, and Active Vision
CountryUnited States
CityBoston, MA
Period2/10/064/10/06

Fingerprint

Non-negative Matrix Factorization
Hyperspectral Image
Parallel algorithms
Source Separation
Factorization
factorization
Source separation
Independent component analysis
matrices
Endmember Extraction
Hyperspectral Imagery
Life sciences
Target Detection
Independent Component Analysis
Distributed Environment
Parallel Computing
Parallel Implementation
Orthogonality
life sciences
Principal Component Analysis

Keywords

  • Blind source separation
  • Hyperspectral images
  • Image processing
  • Linear mixing model
  • Nonnegative matrix factorization
  • Parallel processing

Cite this

Robila, S., & Maciak, L. G. (2006). A parallel unmixing algorithm for hyperspectral images. In Intelligent Robots and Computer Vision XXIV: Algorithms, Techniques, and Active Vision (Vol. 6384). [63840F] https://doi.org/10.1117/12.685655
Robila, Stefan ; Maciak, Lukasz G. / A parallel unmixing algorithm for hyperspectral images. Intelligent Robots and Computer Vision XXIV: Algorithms, Techniques, and Active Vision. Vol. 6384 2006.
@inproceedings{df641817b3fc4cc49f96f8b64b2dec6f,
title = "A parallel unmixing algorithm for hyperspectral images",
abstract = "We present a new algorithm for feature extraction in hyperspectral images based on source separation and parallel computing. In source separation, given a linear mixture of sources, the goal is to recover the components by producing an unmixing matrix. In hyperspectral imagery, the mixing transform and the separated components can be associated with endmembers and their abundances. Source separation based methods have been employed for target detection and classification of hyperspectral images. However, these methods usually involve restrictive conditions on the nature of the results such as orthogonality (in Principal Component Analysis - PCA and Orthogonal Subspace Projection - OSP) of the endmembers or statistical independence (in Independent Component Analysis - ICA) of the abundances nor do they fully satisfy all the conditions included in the Linear Mixing Model. Compared to this, our approach is based on the Nonnegative Matrix Factorization (NMF), a less constraining unmixing method. NMF has the advantage of producing positively defined data, and, with several modifications that we introduce also ensures addition to one. The endmember vectors and the abundances are obtained through a gradient based optimization approach. The algorithm is further modified to run in a parallel environment. The parallel NMF (P-NMF) significantly reduces the time complexity and is shown to also easily port to a distributed environment. Experiments with in-house and Hydice data suggest that NMF outperforms ICA, PCA and OSP for unsupervised endmember extraction. Coupled with its parallel implementation, the new method provides an efficient way for unsupervised unmixing further supporting our efforts in the development of a real time hyperspectral sensing environment with applications to industry and life sciences.",
keywords = "Blind source separation, Hyperspectral images, Image processing, Linear mixing model, Nonnegative matrix factorization, Parallel processing",
author = "Stefan Robila and Maciak, {Lukasz G.}",
year = "2006",
month = "11",
day = "28",
doi = "10.1117/12.685655",
language = "English",
isbn = "0819464821",
volume = "6384",
booktitle = "Intelligent Robots and Computer Vision XXIV",

}

Robila, S & Maciak, LG 2006, A parallel unmixing algorithm for hyperspectral images. in Intelligent Robots and Computer Vision XXIV: Algorithms, Techniques, and Active Vision. vol. 6384, 63840F, Intelligent Robots and Computer Vision XXIV: Algorithms, Techniques, and Active Vision, Boston, MA, United States, 2/10/06. https://doi.org/10.1117/12.685655

A parallel unmixing algorithm for hyperspectral images. / Robila, Stefan; Maciak, Lukasz G.

Intelligent Robots and Computer Vision XXIV: Algorithms, Techniques, and Active Vision. Vol. 6384 2006. 63840F.

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

TY - GEN

T1 - A parallel unmixing algorithm for hyperspectral images

AU - Robila, Stefan

AU - Maciak, Lukasz G.

PY - 2006/11/28

Y1 - 2006/11/28

N2 - We present a new algorithm for feature extraction in hyperspectral images based on source separation and parallel computing. In source separation, given a linear mixture of sources, the goal is to recover the components by producing an unmixing matrix. In hyperspectral imagery, the mixing transform and the separated components can be associated with endmembers and their abundances. Source separation based methods have been employed for target detection and classification of hyperspectral images. However, these methods usually involve restrictive conditions on the nature of the results such as orthogonality (in Principal Component Analysis - PCA and Orthogonal Subspace Projection - OSP) of the endmembers or statistical independence (in Independent Component Analysis - ICA) of the abundances nor do they fully satisfy all the conditions included in the Linear Mixing Model. Compared to this, our approach is based on the Nonnegative Matrix Factorization (NMF), a less constraining unmixing method. NMF has the advantage of producing positively defined data, and, with several modifications that we introduce also ensures addition to one. The endmember vectors and the abundances are obtained through a gradient based optimization approach. The algorithm is further modified to run in a parallel environment. The parallel NMF (P-NMF) significantly reduces the time complexity and is shown to also easily port to a distributed environment. Experiments with in-house and Hydice data suggest that NMF outperforms ICA, PCA and OSP for unsupervised endmember extraction. Coupled with its parallel implementation, the new method provides an efficient way for unsupervised unmixing further supporting our efforts in the development of a real time hyperspectral sensing environment with applications to industry and life sciences.

AB - We present a new algorithm for feature extraction in hyperspectral images based on source separation and parallel computing. In source separation, given a linear mixture of sources, the goal is to recover the components by producing an unmixing matrix. In hyperspectral imagery, the mixing transform and the separated components can be associated with endmembers and their abundances. Source separation based methods have been employed for target detection and classification of hyperspectral images. However, these methods usually involve restrictive conditions on the nature of the results such as orthogonality (in Principal Component Analysis - PCA and Orthogonal Subspace Projection - OSP) of the endmembers or statistical independence (in Independent Component Analysis - ICA) of the abundances nor do they fully satisfy all the conditions included in the Linear Mixing Model. Compared to this, our approach is based on the Nonnegative Matrix Factorization (NMF), a less constraining unmixing method. NMF has the advantage of producing positively defined data, and, with several modifications that we introduce also ensures addition to one. The endmember vectors and the abundances are obtained through a gradient based optimization approach. The algorithm is further modified to run in a parallel environment. The parallel NMF (P-NMF) significantly reduces the time complexity and is shown to also easily port to a distributed environment. Experiments with in-house and Hydice data suggest that NMF outperforms ICA, PCA and OSP for unsupervised endmember extraction. Coupled with its parallel implementation, the new method provides an efficient way for unsupervised unmixing further supporting our efforts in the development of a real time hyperspectral sensing environment with applications to industry and life sciences.

KW - Blind source separation

KW - Hyperspectral images

KW - Image processing

KW - Linear mixing model

KW - Nonnegative matrix factorization

KW - Parallel processing

UR - http://www.scopus.com/inward/record.url?scp=33751300494&partnerID=8YFLogxK

U2 - 10.1117/12.685655

DO - 10.1117/12.685655

M3 - Conference contribution

SN - 0819464821

SN - 9780819464828

VL - 6384

BT - Intelligent Robots and Computer Vision XXIV

ER -

Robila S, Maciak LG. A parallel unmixing algorithm for hyperspectral images. In Intelligent Robots and Computer Vision XXIV: Algorithms, Techniques, and Active Vision. Vol. 6384. 2006. 63840F https://doi.org/10.1117/12.685655