2D and 3D Multimodal Hybrid Face Recognition

    Research output: Contribution to journalArticle

    14 Citations (Scopus)

    Abstract

    We present a 2D and 3D multimodal hybrid face recognition algorithm and demonstrate its performance on the FRGC v1.0 data. We use hybrid (feature-based and holistic) matching for the 3D faces and a holistic matching approach on the 2D faces. Feature-based matching is performed by offline segmenting each 3D face in the gallery into three regions, namely the eyes-forehead, the nose and the cheeks. The cheeks are discarded to avoid facial expressions and hair. During recognition, each feature in the gallery is automatically matched, using a modified ICP algorithm, with a complete probe face. The holistic 3D and 2D face matching is performed using PCA. Individual matching scores are fused after normalization and the results are compared to the BEE baseline performances in order to provide some answers to the first three conjectures of the FRGC. Our multimodal hybrid algorithm substantially outperformed others by achieving 100% verification rate at 0.0006 FAR.
    Original languageEnglish
    Pages (from-to)344-355
    JournalLecture Notes in Computer Science
    Volume3953
    DOIs
    Publication statusPublished - 2006

    Fingerprint

    Face recognition
    Face Recognition
    Face
    Feature Recognition
    Facial Expression
    Recognition Algorithm
    Hybrid Algorithm
    Normalization
    Baseline
    Probe
    Demonstrate

    Cite this

    @article{783794bdbf5f4647b59c8b23250451df,
    title = "2D and 3D Multimodal Hybrid Face Recognition",
    abstract = "We present a 2D and 3D multimodal hybrid face recognition algorithm and demonstrate its performance on the FRGC v1.0 data. We use hybrid (feature-based and holistic) matching for the 3D faces and a holistic matching approach on the 2D faces. Feature-based matching is performed by offline segmenting each 3D face in the gallery into three regions, namely the eyes-forehead, the nose and the cheeks. The cheeks are discarded to avoid facial expressions and hair. During recognition, each feature in the gallery is automatically matched, using a modified ICP algorithm, with a complete probe face. The holistic 3D and 2D face matching is performed using PCA. Individual matching scores are fused after normalization and the results are compared to the BEE baseline performances in order to provide some answers to the first three conjectures of the FRGC. Our multimodal hybrid algorithm substantially outperformed others by achieving 100{\%} verification rate at 0.0006 FAR.",
    author = "Ajmal Mian and Mohammed Bennamoun and Robyn Owens",
    year = "2006",
    doi = "10.1007/11744078_27",
    language = "English",
    volume = "3953",
    pages = "344--355",
    journal = "Lecture Notes in Computer Science",
    issn = "0302-9743",
    publisher = "Springer-Verlag London Ltd.",

    }

    2D and 3D Multimodal Hybrid Face Recognition. / Mian, Ajmal; Bennamoun, Mohammed; Owens, Robyn.

    In: Lecture Notes in Computer Science, Vol. 3953, 2006, p. 344-355.

    Research output: Contribution to journalArticle

    TY - JOUR

    T1 - 2D and 3D Multimodal Hybrid Face Recognition

    AU - Mian, Ajmal

    AU - Bennamoun, Mohammed

    AU - Owens, Robyn

    PY - 2006

    Y1 - 2006

    N2 - We present a 2D and 3D multimodal hybrid face recognition algorithm and demonstrate its performance on the FRGC v1.0 data. We use hybrid (feature-based and holistic) matching for the 3D faces and a holistic matching approach on the 2D faces. Feature-based matching is performed by offline segmenting each 3D face in the gallery into three regions, namely the eyes-forehead, the nose and the cheeks. The cheeks are discarded to avoid facial expressions and hair. During recognition, each feature in the gallery is automatically matched, using a modified ICP algorithm, with a complete probe face. The holistic 3D and 2D face matching is performed using PCA. Individual matching scores are fused after normalization and the results are compared to the BEE baseline performances in order to provide some answers to the first three conjectures of the FRGC. Our multimodal hybrid algorithm substantially outperformed others by achieving 100% verification rate at 0.0006 FAR.

    AB - We present a 2D and 3D multimodal hybrid face recognition algorithm and demonstrate its performance on the FRGC v1.0 data. We use hybrid (feature-based and holistic) matching for the 3D faces and a holistic matching approach on the 2D faces. Feature-based matching is performed by offline segmenting each 3D face in the gallery into three regions, namely the eyes-forehead, the nose and the cheeks. The cheeks are discarded to avoid facial expressions and hair. During recognition, each feature in the gallery is automatically matched, using a modified ICP algorithm, with a complete probe face. The holistic 3D and 2D face matching is performed using PCA. Individual matching scores are fused after normalization and the results are compared to the BEE baseline performances in order to provide some answers to the first three conjectures of the FRGC. Our multimodal hybrid algorithm substantially outperformed others by achieving 100% verification rate at 0.0006 FAR.

    U2 - 10.1007/11744078_27

    DO - 10.1007/11744078_27

    M3 - Article

    VL - 3953

    SP - 344

    EP - 355

    JO - Lecture Notes in Computer Science

    JF - Lecture Notes in Computer Science

    SN - 0302-9743

    ER -