Facial Feature Extraction on Fiducial Points and Used in Face Recognition

One of the main issues of face recognition is to decide what features to represent a face. In this paper, we present a new algorithm that extracts facial features on some fiducial points. 17 fiducial points are automatically located by Active Appearance Models (AAMs) and characterized with Gabor wavelet analysis. The features are evaluated by a face database, which includes more than 400 images of 40 persons. Experimental results show the algorithm is effective and robust under some conditions.