Exploiting sparsity in hyperspectral image classification via graphical models

Umamahesh Srinivas, Yi Chen, Vishal Monga, Nasser M. Nasrabadi, Trac D. Tran

    Research output: Contribution to journalArticle

    59 Citations (Scopus)

    Abstract

    A significant recent advance in hyperspectral image (HSI) classification relies on the observation that the spectral signature of a pixel can be represented by a sparse linear combination of training spectra from an overcomplete dictionary. A spatiospectral notion of sparsity is further captured by developing a joint sparsity model, wherein spectral signatures of pixels in a local spatial neighborhood (of the pixel of interest) are constrained to be represented by a common collection of training spectra, albeit with different weights. A challenging open problem is to effectively capture the class conditional correlations between these multiple sparse representations corresponding to different pixels in the spatial neighborhood. We propose a probabilistic graphical model framework to explicitly mine the conditional dependences between these distinct sparse features. Our graphical models are synthesized using simple tree structures which can be discriminatively learnt (even with limited training samples) for classification. Experiments on benchmark HSI data sets reveal significant improvements over existing approaches in classification rates as well as robustness to choice of training.

    Original languageEnglish (US)
    Article number6297997
    Pages (from-to)505-509
    Number of pages5
    JournalIEEE Geoscience and Remote Sensing Letters
    Volume10
    Issue number3
    DOIs
    StatePublished - Jan 1 2013

    Fingerprint

    Image classification
    image classification
    pixel
    Pixels
    Glossaries
    experiment
    Experiments

    All Science Journal Classification (ASJC) codes

    • Geotechnical Engineering and Engineering Geology
    • Electrical and Electronic Engineering

    Cite this

    Srinivas, Umamahesh ; Chen, Yi ; Monga, Vishal ; Nasrabadi, Nasser M. ; Tran, Trac D. / Exploiting sparsity in hyperspectral image classification via graphical models. In: IEEE Geoscience and Remote Sensing Letters. 2013 ; Vol. 10, No. 3. pp. 505-509.
    @article{698d00c575264c98b2512001c9db47ee,
    title = "Exploiting sparsity in hyperspectral image classification via graphical models",
    abstract = "A significant recent advance in hyperspectral image (HSI) classification relies on the observation that the spectral signature of a pixel can be represented by a sparse linear combination of training spectra from an overcomplete dictionary. A spatiospectral notion of sparsity is further captured by developing a joint sparsity model, wherein spectral signatures of pixels in a local spatial neighborhood (of the pixel of interest) are constrained to be represented by a common collection of training spectra, albeit with different weights. A challenging open problem is to effectively capture the class conditional correlations between these multiple sparse representations corresponding to different pixels in the spatial neighborhood. We propose a probabilistic graphical model framework to explicitly mine the conditional dependences between these distinct sparse features. Our graphical models are synthesized using simple tree structures which can be discriminatively learnt (even with limited training samples) for classification. Experiments on benchmark HSI data sets reveal significant improvements over existing approaches in classification rates as well as robustness to choice of training.",
    author = "Umamahesh Srinivas and Yi Chen and Vishal Monga and Nasrabadi, {Nasser M.} and Tran, {Trac D.}",
    year = "2013",
    month = "1",
    day = "1",
    doi = "10.1109/LGRS.2012.2211858",
    language = "English (US)",
    volume = "10",
    pages = "505--509",
    journal = "IEEE Geoscience and Remote Sensing Letters",
    issn = "1545-598X",
    publisher = "Institute of Electrical and Electronics Engineers Inc.",
    number = "3",

    }

    Exploiting sparsity in hyperspectral image classification via graphical models. / Srinivas, Umamahesh; Chen, Yi; Monga, Vishal; Nasrabadi, Nasser M.; Tran, Trac D.

    In: IEEE Geoscience and Remote Sensing Letters, Vol. 10, No. 3, 6297997, 01.01.2013, p. 505-509.

    Research output: Contribution to journalArticle

    TY - JOUR

    T1 - Exploiting sparsity in hyperspectral image classification via graphical models

    AU - Srinivas, Umamahesh

    AU - Chen, Yi

    AU - Monga, Vishal

    AU - Nasrabadi, Nasser M.

    AU - Tran, Trac D.

    PY - 2013/1/1

    Y1 - 2013/1/1

    N2 - A significant recent advance in hyperspectral image (HSI) classification relies on the observation that the spectral signature of a pixel can be represented by a sparse linear combination of training spectra from an overcomplete dictionary. A spatiospectral notion of sparsity is further captured by developing a joint sparsity model, wherein spectral signatures of pixels in a local spatial neighborhood (of the pixel of interest) are constrained to be represented by a common collection of training spectra, albeit with different weights. A challenging open problem is to effectively capture the class conditional correlations between these multiple sparse representations corresponding to different pixels in the spatial neighborhood. We propose a probabilistic graphical model framework to explicitly mine the conditional dependences between these distinct sparse features. Our graphical models are synthesized using simple tree structures which can be discriminatively learnt (even with limited training samples) for classification. Experiments on benchmark HSI data sets reveal significant improvements over existing approaches in classification rates as well as robustness to choice of training.

    AB - A significant recent advance in hyperspectral image (HSI) classification relies on the observation that the spectral signature of a pixel can be represented by a sparse linear combination of training spectra from an overcomplete dictionary. A spatiospectral notion of sparsity is further captured by developing a joint sparsity model, wherein spectral signatures of pixels in a local spatial neighborhood (of the pixel of interest) are constrained to be represented by a common collection of training spectra, albeit with different weights. A challenging open problem is to effectively capture the class conditional correlations between these multiple sparse representations corresponding to different pixels in the spatial neighborhood. We propose a probabilistic graphical model framework to explicitly mine the conditional dependences between these distinct sparse features. Our graphical models are synthesized using simple tree structures which can be discriminatively learnt (even with limited training samples) for classification. Experiments on benchmark HSI data sets reveal significant improvements over existing approaches in classification rates as well as robustness to choice of training.

    UR - http://www.scopus.com/inward/record.url?scp=84870541423&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84870541423&partnerID=8YFLogxK

    U2 - 10.1109/LGRS.2012.2211858

    DO - 10.1109/LGRS.2012.2211858

    M3 - Article

    AN - SCOPUS:84870541423

    VL - 10

    SP - 505

    EP - 509

    JO - IEEE Geoscience and Remote Sensing Letters

    JF - IEEE Geoscience and Remote Sensing Letters

    SN - 1545-598X

    IS - 3

    M1 - 6297997

    ER -