Linear spectral unmixing using collaborative sparse regression and correlated supports

Yoann Altmann, Marcelo Pereyra, Jose Bioucas Dias

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper presents a new Bayesian collaborative sparse regression method for linear unmixing of hyperspectral images. Our contribution is twofold; first, we propose a new Bayesian model for structured sparse regression in which the supports of the sparse abundance vectors are a priori spatially correlated across pixels. Secondly, we propose an advanced Markov chain Monte Carlo algorithm to estimate the posterior probabilities that materials are present or absent in each pixel, and, conditionally to the maximum marginal a posteriori configuration of this support, compute the MMSE estimates of the abundance vectors. A remarkable property of this algorithm is that it self-adjusts the values of the parameters of the Markov random field, thus relieving practitioners from setting regularisation parameters, namely by cross-validation. The proposed methodology is illustrated with real hyperspectral data.

Original languageEnglish
Title of host publication2015 7th Workshop on Hyperspectral Image and Signal Processing
Subtitle of host publicationEvolution in Remote Sensing (WHISPERS)
PublisherIEEE
ISBN (Electronic)9781467390156
DOIs
Publication statusPublished - 23 Oct 2017
Event7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing 2015 - Tokyo, Japan
Duration: 2 Jun 20155 Jun 2015

Publication series

NameWorkshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)
PublisherIEEE
ISSN (Electronic)2158-6276

Conference

Conference7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing 2015
Abbreviated titleWHISPERS 2015
CountryJapan
CityTokyo
Period2/06/155/06/15

Fingerprint

Pixels
Markov processes

Keywords

  • Bayesian estimation
  • Collaborative sparse regression
  • Markov chain Monte Carlo methods
  • Markov random fields
  • Spectral unmixing

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Signal Processing

Cite this

Altmann, Y., Pereyra, M., & Dias, J. B. (2017). Linear spectral unmixing using collaborative sparse regression and correlated supports. In 2015 7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS) [8075438] (Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)). IEEE. https://doi.org/10.1109/WHISPERS.2015.8075438
Altmann, Yoann ; Pereyra, Marcelo ; Dias, Jose Bioucas. / Linear spectral unmixing using collaborative sparse regression and correlated supports. 2015 7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS). IEEE, 2017. (Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)).
@inproceedings{f489735cbff744c9b112cf647325da1c,
title = "Linear spectral unmixing using collaborative sparse regression and correlated supports",
abstract = "This paper presents a new Bayesian collaborative sparse regression method for linear unmixing of hyperspectral images. Our contribution is twofold; first, we propose a new Bayesian model for structured sparse regression in which the supports of the sparse abundance vectors are a priori spatially correlated across pixels. Secondly, we propose an advanced Markov chain Monte Carlo algorithm to estimate the posterior probabilities that materials are present or absent in each pixel, and, conditionally to the maximum marginal a posteriori configuration of this support, compute the MMSE estimates of the abundance vectors. A remarkable property of this algorithm is that it self-adjusts the values of the parameters of the Markov random field, thus relieving practitioners from setting regularisation parameters, namely by cross-validation. The proposed methodology is illustrated with real hyperspectral data.",
keywords = "Bayesian estimation, Collaborative sparse regression, Markov chain Monte Carlo methods, Markov random fields, Spectral unmixing",
author = "Yoann Altmann and Marcelo Pereyra and Dias, {Jose Bioucas}",
year = "2017",
month = "10",
day = "23",
doi = "10.1109/WHISPERS.2015.8075438",
language = "English",
series = "Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)",
publisher = "IEEE",
booktitle = "2015 7th Workshop on Hyperspectral Image and Signal Processing",
address = "United States",

}

Altmann, Y, Pereyra, M & Dias, JB 2017, Linear spectral unmixing using collaborative sparse regression and correlated supports. in 2015 7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)., 8075438, Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS), IEEE, 7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing 2015, Tokyo, Japan, 2/06/15. https://doi.org/10.1109/WHISPERS.2015.8075438

Linear spectral unmixing using collaborative sparse regression and correlated supports. / Altmann, Yoann; Pereyra, Marcelo; Dias, Jose Bioucas.

2015 7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS). IEEE, 2017. 8075438 (Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Linear spectral unmixing using collaborative sparse regression and correlated supports

AU - Altmann, Yoann

AU - Pereyra, Marcelo

AU - Dias, Jose Bioucas

PY - 2017/10/23

Y1 - 2017/10/23

N2 - This paper presents a new Bayesian collaborative sparse regression method for linear unmixing of hyperspectral images. Our contribution is twofold; first, we propose a new Bayesian model for structured sparse regression in which the supports of the sparse abundance vectors are a priori spatially correlated across pixels. Secondly, we propose an advanced Markov chain Monte Carlo algorithm to estimate the posterior probabilities that materials are present or absent in each pixel, and, conditionally to the maximum marginal a posteriori configuration of this support, compute the MMSE estimates of the abundance vectors. A remarkable property of this algorithm is that it self-adjusts the values of the parameters of the Markov random field, thus relieving practitioners from setting regularisation parameters, namely by cross-validation. The proposed methodology is illustrated with real hyperspectral data.

AB - This paper presents a new Bayesian collaborative sparse regression method for linear unmixing of hyperspectral images. Our contribution is twofold; first, we propose a new Bayesian model for structured sparse regression in which the supports of the sparse abundance vectors are a priori spatially correlated across pixels. Secondly, we propose an advanced Markov chain Monte Carlo algorithm to estimate the posterior probabilities that materials are present or absent in each pixel, and, conditionally to the maximum marginal a posteriori configuration of this support, compute the MMSE estimates of the abundance vectors. A remarkable property of this algorithm is that it self-adjusts the values of the parameters of the Markov random field, thus relieving practitioners from setting regularisation parameters, namely by cross-validation. The proposed methodology is illustrated with real hyperspectral data.

KW - Bayesian estimation

KW - Collaborative sparse regression

KW - Markov chain Monte Carlo methods

KW - Markov random fields

KW - Spectral unmixing

UR - http://www.scopus.com/inward/record.url?scp=85039167056&partnerID=8YFLogxK

U2 - 10.1109/WHISPERS.2015.8075438

DO - 10.1109/WHISPERS.2015.8075438

M3 - Conference contribution

T3 - Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)

BT - 2015 7th Workshop on Hyperspectral Image and Signal Processing

PB - IEEE

ER -

Altmann Y, Pereyra M, Dias JB. Linear spectral unmixing using collaborative sparse regression and correlated supports. In 2015 7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS). IEEE. 2017. 8075438. (Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)). https://doi.org/10.1109/WHISPERS.2015.8075438