Holistic feature extraction for automatic image annotation

Supheakmungkol Sarin, Michael Fahrmair, Matthias Wagner, Wataru Kameyama

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

    4 Citations (Scopus)

    Abstract

    Automating the annotation process of digital images is a crucial step towards efficient and effective management of this increasingly high volume of content. It is, nevertheless, an extremely challenging task for the research community. One of the main bottle necks is the lack of integrity and diversity of features. We solve this problem by proposing to utilize 43 image features that cover the holistic content of the image from global to subject, background, and scene. In our approach, saliency regions and background are separated without prior knowledge. Each of them together with the whole image is treated independently for feature extraction. Extensive experiments were designed to show the efficiency and effectiveness of our approach. We chose two publicly available datasets manually annotated and with the diverse nature of images for our experiments, namely, the Corel5k and ESP Game datasets. They contain 5,000 images with 260 keywords and 20,770 images with 268 keywords, respectively. Through empirical experiments, it is confirmed that by using our features with the state-of-the-art technique, we achieve superior performance in many metrics, particularly in auto-annotation.

    Original languageEnglish
    Title of host publicationProceedings of the 2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011
    Pages59-66
    Number of pages8
    DOIs
    Publication statusPublished - 2011
    Event2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011 - Loutraki
    Duration: 2011 Jun 282011 Jun 30

    Other

    Other2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011
    CityLoutraki
    Period11/6/2811/6/30

    Fingerprint

    Feature extraction
    Experiments
    Bottles

    Keywords

    • Automatic image annotation
    • Background
    • Holistic feature extraction
    • K nearest neighbours
    • Saliency regions

    ASJC Scopus subject areas

    • Computer Graphics and Computer-Aided Design
    • Computer Vision and Pattern Recognition

    Cite this

    Sarin, S., Fahrmair, M., Wagner, M., & Kameyama, W. (2011). Holistic feature extraction for automatic image annotation. In Proceedings of the 2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011 (pp. 59-66). [5992172] https://doi.org/10.1109/MUE.2011.22

    Holistic feature extraction for automatic image annotation. / Sarin, Supheakmungkol; Fahrmair, Michael; Wagner, Matthias; Kameyama, Wataru.

    Proceedings of the 2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011. 2011. p. 59-66 5992172.

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

    Sarin, S, Fahrmair, M, Wagner, M & Kameyama, W 2011, Holistic feature extraction for automatic image annotation. in Proceedings of the 2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011., 5992172, pp. 59-66, 2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011, Loutraki, 11/6/28. https://doi.org/10.1109/MUE.2011.22
    Sarin S, Fahrmair M, Wagner M, Kameyama W. Holistic feature extraction for automatic image annotation. In Proceedings of the 2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011. 2011. p. 59-66. 5992172 https://doi.org/10.1109/MUE.2011.22
    Sarin, Supheakmungkol ; Fahrmair, Michael ; Wagner, Matthias ; Kameyama, Wataru. / Holistic feature extraction for automatic image annotation. Proceedings of the 2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011. 2011. pp. 59-66
    @inproceedings{e208eaac1c4b44daacea6bdcb93c79eb,
    title = "Holistic feature extraction for automatic image annotation",
    abstract = "Automating the annotation process of digital images is a crucial step towards efficient and effective management of this increasingly high volume of content. It is, nevertheless, an extremely challenging task for the research community. One of the main bottle necks is the lack of integrity and diversity of features. We solve this problem by proposing to utilize 43 image features that cover the holistic content of the image from global to subject, background, and scene. In our approach, saliency regions and background are separated without prior knowledge. Each of them together with the whole image is treated independently for feature extraction. Extensive experiments were designed to show the efficiency and effectiveness of our approach. We chose two publicly available datasets manually annotated and with the diverse nature of images for our experiments, namely, the Corel5k and ESP Game datasets. They contain 5,000 images with 260 keywords and 20,770 images with 268 keywords, respectively. Through empirical experiments, it is confirmed that by using our features with the state-of-the-art technique, we achieve superior performance in many metrics, particularly in auto-annotation.",
    keywords = "Automatic image annotation, Background, Holistic feature extraction, K nearest neighbours, Saliency regions",
    author = "Supheakmungkol Sarin and Michael Fahrmair and Matthias Wagner and Wataru Kameyama",
    year = "2011",
    doi = "10.1109/MUE.2011.22",
    language = "English",
    isbn = "9780769544700",
    pages = "59--66",
    booktitle = "Proceedings of the 2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011",

    }

    TY - GEN

    T1 - Holistic feature extraction for automatic image annotation

    AU - Sarin, Supheakmungkol

    AU - Fahrmair, Michael

    AU - Wagner, Matthias

    AU - Kameyama, Wataru

    PY - 2011

    Y1 - 2011

    N2 - Automating the annotation process of digital images is a crucial step towards efficient and effective management of this increasingly high volume of content. It is, nevertheless, an extremely challenging task for the research community. One of the main bottle necks is the lack of integrity and diversity of features. We solve this problem by proposing to utilize 43 image features that cover the holistic content of the image from global to subject, background, and scene. In our approach, saliency regions and background are separated without prior knowledge. Each of them together with the whole image is treated independently for feature extraction. Extensive experiments were designed to show the efficiency and effectiveness of our approach. We chose two publicly available datasets manually annotated and with the diverse nature of images for our experiments, namely, the Corel5k and ESP Game datasets. They contain 5,000 images with 260 keywords and 20,770 images with 268 keywords, respectively. Through empirical experiments, it is confirmed that by using our features with the state-of-the-art technique, we achieve superior performance in many metrics, particularly in auto-annotation.

    AB - Automating the annotation process of digital images is a crucial step towards efficient and effective management of this increasingly high volume of content. It is, nevertheless, an extremely challenging task for the research community. One of the main bottle necks is the lack of integrity and diversity of features. We solve this problem by proposing to utilize 43 image features that cover the holistic content of the image from global to subject, background, and scene. In our approach, saliency regions and background are separated without prior knowledge. Each of them together with the whole image is treated independently for feature extraction. Extensive experiments were designed to show the efficiency and effectiveness of our approach. We chose two publicly available datasets manually annotated and with the diverse nature of images for our experiments, namely, the Corel5k and ESP Game datasets. They contain 5,000 images with 260 keywords and 20,770 images with 268 keywords, respectively. Through empirical experiments, it is confirmed that by using our features with the state-of-the-art technique, we achieve superior performance in many metrics, particularly in auto-annotation.

    KW - Automatic image annotation

    KW - Background

    KW - Holistic feature extraction

    KW - K nearest neighbours

    KW - Saliency regions

    UR - http://www.scopus.com/inward/record.url?scp=80052595697&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=80052595697&partnerID=8YFLogxK

    U2 - 10.1109/MUE.2011.22

    DO - 10.1109/MUE.2011.22

    M3 - Conference contribution

    AN - SCOPUS:80052595697

    SN - 9780769544700

    SP - 59

    EP - 66

    BT - Proceedings of the 2011 5th FTRA International Conference on Multimedia and Ubiquitous Engineering, MUE 2011

    ER -