Unsupervised post-nonlinear unmixing of hyperspectral images using a hamiltonian Monte Carlo algorithm

Yoann Altmann, Nicolas Dobigeon, Jean-Yves Tourneret

Research output: Contribution to journalArticle

Abstract

This paper presents a nonlinear mixing model for hyperspectral image unmixing. The proposed model assumes that the pixel reflectances are post-nonlinear functions of unknown pure spectral components contaminated by an additive white Gaussian noise. These nonlinear functions are approximated using second-order polynomials leading to a polynomial postnonlinear mixing model. A Bayesian algorithm is proposed to estimate the parameters involved in the model yielding an unsupervised nonlinear unmixing algorithm. Due to the large number of parameters to be estimated, an efficient Hamiltonian Monte Carlo algorithm is investigated. The classical leapfrog steps of this algorithm are modified to handle the parameter constraints. The performance of the unmixing strategy, including convergence and parameter tuning, is first evaluated on synthetic data. Simulations conducted with real data finally show the accuracy of the proposed unmixing strategy for the analysis of hyperspectral images.

Original languageEnglish
Article number6778790
Pages (from-to)2663-2675
Number of pages13
JournalIEEE Transactions on Image Processing
Volume23
Issue number6
DOIs
Publication statusPublished - Jun 2014

Fingerprint

Hamiltonians
Polynomials
Tuning
Pixels

Keywords

  • Hamiltonian Monte Carlo
  • Hyperspectral imagery
  • Post-nonlinear model
  • Unsupervised spectral unmixing

ASJC Scopus subject areas

  • Software
  • Computer Graphics and Computer-Aided Design

Cite this

@article{98b811030285483fb95231cfc920f8d3,
title = "Unsupervised post-nonlinear unmixing of hyperspectral images using a hamiltonian Monte Carlo algorithm",
abstract = "This paper presents a nonlinear mixing model for hyperspectral image unmixing. The proposed model assumes that the pixel reflectances are post-nonlinear functions of unknown pure spectral components contaminated by an additive white Gaussian noise. These nonlinear functions are approximated using second-order polynomials leading to a polynomial postnonlinear mixing model. A Bayesian algorithm is proposed to estimate the parameters involved in the model yielding an unsupervised nonlinear unmixing algorithm. Due to the large number of parameters to be estimated, an efficient Hamiltonian Monte Carlo algorithm is investigated. The classical leapfrog steps of this algorithm are modified to handle the parameter constraints. The performance of the unmixing strategy, including convergence and parameter tuning, is first evaluated on synthetic data. Simulations conducted with real data finally show the accuracy of the proposed unmixing strategy for the analysis of hyperspectral images.",
keywords = "Hamiltonian Monte Carlo, Hyperspectral imagery, Post-nonlinear model, Unsupervised spectral unmixing",
author = "Yoann Altmann and Nicolas Dobigeon and Jean-Yves Tourneret",
year = "2014",
month = "6",
doi = "10.1109/TIP.2014.2314022",
language = "English",
volume = "23",
pages = "2663--2675",
journal = "IEEE Transactions on Image Processing",
issn = "1057-7149",
publisher = "IEEE",
number = "6",

}

Unsupervised post-nonlinear unmixing of hyperspectral images using a hamiltonian Monte Carlo algorithm. / Altmann, Yoann; Dobigeon, Nicolas; Tourneret, Jean-Yves.

In: IEEE Transactions on Image Processing, Vol. 23, No. 6, 6778790, 06.2014, p. 2663-2675.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Unsupervised post-nonlinear unmixing of hyperspectral images using a hamiltonian Monte Carlo algorithm

AU - Altmann, Yoann

AU - Dobigeon, Nicolas

AU - Tourneret, Jean-Yves

PY - 2014/6

Y1 - 2014/6

N2 - This paper presents a nonlinear mixing model for hyperspectral image unmixing. The proposed model assumes that the pixel reflectances are post-nonlinear functions of unknown pure spectral components contaminated by an additive white Gaussian noise. These nonlinear functions are approximated using second-order polynomials leading to a polynomial postnonlinear mixing model. A Bayesian algorithm is proposed to estimate the parameters involved in the model yielding an unsupervised nonlinear unmixing algorithm. Due to the large number of parameters to be estimated, an efficient Hamiltonian Monte Carlo algorithm is investigated. The classical leapfrog steps of this algorithm are modified to handle the parameter constraints. The performance of the unmixing strategy, including convergence and parameter tuning, is first evaluated on synthetic data. Simulations conducted with real data finally show the accuracy of the proposed unmixing strategy for the analysis of hyperspectral images.

AB - This paper presents a nonlinear mixing model for hyperspectral image unmixing. The proposed model assumes that the pixel reflectances are post-nonlinear functions of unknown pure spectral components contaminated by an additive white Gaussian noise. These nonlinear functions are approximated using second-order polynomials leading to a polynomial postnonlinear mixing model. A Bayesian algorithm is proposed to estimate the parameters involved in the model yielding an unsupervised nonlinear unmixing algorithm. Due to the large number of parameters to be estimated, an efficient Hamiltonian Monte Carlo algorithm is investigated. The classical leapfrog steps of this algorithm are modified to handle the parameter constraints. The performance of the unmixing strategy, including convergence and parameter tuning, is first evaluated on synthetic data. Simulations conducted with real data finally show the accuracy of the proposed unmixing strategy for the analysis of hyperspectral images.

KW - Hamiltonian Monte Carlo

KW - Hyperspectral imagery

KW - Post-nonlinear model

KW - Unsupervised spectral unmixing

UR - http://www.scopus.com/inward/record.url?scp=84903793428&partnerID=8YFLogxK

U2 - 10.1109/TIP.2014.2314022

DO - 10.1109/TIP.2014.2314022

M3 - Article

C2 - 24686282

AN - SCOPUS:84903793428

VL - 23

SP - 2663

EP - 2675

JO - IEEE Transactions on Image Processing

JF - IEEE Transactions on Image Processing

SN - 1057-7149

IS - 6

M1 - 6778790

ER -