Subsribe to our RSS Feed

Machine Learning

Learning based Big Image Data Analysis

  • Overview


    For learning based image recognition, we adopt the Bayesian model and dimensional reduction algorithms. Variations of a face are considered by probabilistic function and similarities of two images are computed by using them. Contrary to previous works, we consider intra personal variation in diverse way.


  • Research Subjects


    Bayesian model for image recognition


    Supervised way
    -Intra personal variations are modeled as probabilistic functions in a supervised way.


    Unsupervised way
    -Intra personal variations are grouped by unsupervised clustering.
    Feature transform and learning


    Feature transformation using tree structure.


    The feature transformation for face recognition.
    Rotationally invariant feature encoding


    Rotationally invariant feature encoding is very important for image representation.
    Sorted consecutive technic and codebook learning is used for the feature encoding.



  • Contents of the Research


    1. Image detection

    2. Image description

    3. Noise reduction / Illumination compensation



  • Required Knowledge for the Research


    • Probability and Statistics

    • Linear Algebra

    • Pattern Recognition

    • Image Processing



  • Reference works


    [1] S. Lee, S. Choi and H. S Yang, Bag-of-binary-features for fast image representation, Electronics Letters, 2015-04

    [2] J. Ryu, S. Hong and H. S Yang, Sorted Consecutive Local Binary Pattern for Texture Classification, Image Processing, IEEE Transactions on, 2015-04

    [3] B. Moghaddam, T. Jebara, and A. Pentland, Bayesian face recognition, Pattern Recognition, 2000