Structured Sparse Priors for Image Classification

Umamahesh Srinivas, Yuanming Suo, Minh Dao, Vishal Monga, Trac D. Tran

    Research output: Contribution to journalArticle

    28 Citations (Scopus)

    Abstract

    Model-based compressive sensing (CS) exploits the structure inherent in sparse signals for the design of better signal recovery algorithms. This information about structure is often captured in the form of a prior on the sparse coefficients, with the Laplacian being the most common such choice (leading to l1-norm minimization). Recent work has exploited the discriminative capability of sparse representations for image classification by employing class-specific dictionaries in the CS framework. Our contribution is a logical extension of these ideas into structured sparsity for classification. We introduce the notion of discriminative class-specific priors in conjunction with class specific dictionaries, specifically the spike-and-slab prior widely applied in Bayesian sparse regression. Significantly, the proposed framework takes the burden off the demand for abundant training image samples necessary for the success of sparsity-based classification schemes. We demonstrate this practical benefit of our approach in important applications, such as face recognition and object categorization.

    Original languageEnglish (US)
    Article number7055925
    Pages (from-to)1763-1776
    Number of pages14
    JournalIEEE Transactions on Image Processing
    Volume24
    Issue number6
    DOIs
    StatePublished - Jun 1 2015

    Fingerprint

    Image classification
    Glossaries
    Face recognition
    Recovery

    All Science Journal Classification (ASJC) codes

    • Software
    • Computer Graphics and Computer-Aided Design

    Cite this

    Srinivas, Umamahesh ; Suo, Yuanming ; Dao, Minh ; Monga, Vishal ; Tran, Trac D. / Structured Sparse Priors for Image Classification. In: IEEE Transactions on Image Processing. 2015 ; Vol. 24, No. 6. pp. 1763-1776.
    @article{b77563b057884f76b4a09d1fc316c95c,
    title = "Structured Sparse Priors for Image Classification",
    abstract = "Model-based compressive sensing (CS) exploits the structure inherent in sparse signals for the design of better signal recovery algorithms. This information about structure is often captured in the form of a prior on the sparse coefficients, with the Laplacian being the most common such choice (leading to l1-norm minimization). Recent work has exploited the discriminative capability of sparse representations for image classification by employing class-specific dictionaries in the CS framework. Our contribution is a logical extension of these ideas into structured sparsity for classification. We introduce the notion of discriminative class-specific priors in conjunction with class specific dictionaries, specifically the spike-and-slab prior widely applied in Bayesian sparse regression. Significantly, the proposed framework takes the burden off the demand for abundant training image samples necessary for the success of sparsity-based classification schemes. We demonstrate this practical benefit of our approach in important applications, such as face recognition and object categorization.",
    author = "Umamahesh Srinivas and Yuanming Suo and Minh Dao and Vishal Monga and Tran, {Trac D.}",
    year = "2015",
    month = "6",
    day = "1",
    doi = "10.1109/TIP.2015.2409572",
    language = "English (US)",
    volume = "24",
    pages = "1763--1776",
    journal = "IEEE Transactions on Image Processing",
    issn = "1057-7149",
    publisher = "Institute of Electrical and Electronics Engineers Inc.",
    number = "6",

    }

    Srinivas, U, Suo, Y, Dao, M, Monga, V & Tran, TD 2015, 'Structured Sparse Priors for Image Classification', IEEE Transactions on Image Processing, vol. 24, no. 6, 7055925, pp. 1763-1776. https://doi.org/10.1109/TIP.2015.2409572

    Structured Sparse Priors for Image Classification. / Srinivas, Umamahesh; Suo, Yuanming; Dao, Minh; Monga, Vishal; Tran, Trac D.

    In: IEEE Transactions on Image Processing, Vol. 24, No. 6, 7055925, 01.06.2015, p. 1763-1776.

    Research output: Contribution to journalArticle

    TY - JOUR

    T1 - Structured Sparse Priors for Image Classification

    AU - Srinivas, Umamahesh

    AU - Suo, Yuanming

    AU - Dao, Minh

    AU - Monga, Vishal

    AU - Tran, Trac D.

    PY - 2015/6/1

    Y1 - 2015/6/1

    N2 - Model-based compressive sensing (CS) exploits the structure inherent in sparse signals for the design of better signal recovery algorithms. This information about structure is often captured in the form of a prior on the sparse coefficients, with the Laplacian being the most common such choice (leading to l1-norm minimization). Recent work has exploited the discriminative capability of sparse representations for image classification by employing class-specific dictionaries in the CS framework. Our contribution is a logical extension of these ideas into structured sparsity for classification. We introduce the notion of discriminative class-specific priors in conjunction with class specific dictionaries, specifically the spike-and-slab prior widely applied in Bayesian sparse regression. Significantly, the proposed framework takes the burden off the demand for abundant training image samples necessary for the success of sparsity-based classification schemes. We demonstrate this practical benefit of our approach in important applications, such as face recognition and object categorization.

    AB - Model-based compressive sensing (CS) exploits the structure inherent in sparse signals for the design of better signal recovery algorithms. This information about structure is often captured in the form of a prior on the sparse coefficients, with the Laplacian being the most common such choice (leading to l1-norm minimization). Recent work has exploited the discriminative capability of sparse representations for image classification by employing class-specific dictionaries in the CS framework. Our contribution is a logical extension of these ideas into structured sparsity for classification. We introduce the notion of discriminative class-specific priors in conjunction with class specific dictionaries, specifically the spike-and-slab prior widely applied in Bayesian sparse regression. Significantly, the proposed framework takes the burden off the demand for abundant training image samples necessary for the success of sparsity-based classification schemes. We demonstrate this practical benefit of our approach in important applications, such as face recognition and object categorization.

    UR - http://www.scopus.com/inward/record.url?scp=84926428948&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84926428948&partnerID=8YFLogxK

    U2 - 10.1109/TIP.2015.2409572

    DO - 10.1109/TIP.2015.2409572

    M3 - Article

    C2 - 25769162

    AN - SCOPUS:84926428948

    VL - 24

    SP - 1763

    EP - 1776

    JO - IEEE Transactions on Image Processing

    JF - IEEE Transactions on Image Processing

    SN - 1057-7149

    IS - 6

    M1 - 7055925

    ER -