Hybrid component-based face recognition
Amanda Gumede, Serestina Viriri, Mandlenkosi Victor Gwetu · 2017
In this paper a hybrid component-based approach for facial recognition is proposed, as an alternative to current facial recognition techniques. The algorithm detects the facial components (eyes, mouth, nose), and then extract textural and shape features. If the face is partially occluded, the successfully detected components from the non-occluded facial part are used for recognition. The feature descriptors, Garbo Filters and Zernike Moments have been used for textural and shape features, respectively. The combination of these two descriptors has a significant high discriminative level of extracting distinctive features from the facial components. The experiments were carried out on four different facial databases, the ORL, FERET, FEI and CMU. The experimental results achieved an overall accuracy rate of 93.9%. The experimental results show that component-based facial recognition is effective.