Abstract：In this paper, a robust hierarchical approach based on AdaBoost algorithm and C-V method was presented for facial features localization. First, four kinds of detectors were trained by AdaBoost algorithm for detecting faces, eyes, noses and mouths. Second, face regions were detected using the face detector combined with a rule of face edge. Third, the eye regions, nose regions and mouth regions were detected using the facial feature detectors, and the feature contours and feature points were extracted from the feature regions by C-V method. The experiments on DTU_IMM face test set resulted in 94.6% accuracy rate on eyes, 95.3% on noses and 98.4% on mouths, and the positions of the extracted feature points were accurate. Results show that the proposed approach is efficient and robust.
