Extraction of Face Features Using Various Techniques

Abstract

This thesis aims at devising a novel method of feature extraction of face images which proves to be faster and more accurate than the existing methods defined by wavelet, curvelet and ridgelet transforms. DOST method of extracting features from face images keeps into account every minute detail of the face image i.e both spatial and frequency based features. The application of LDA method onto the DOST features in order to reduce the dimensionality of the method further helps in making the process of feature extraction faster and hence reduces the time complexity of the feature extraction method. The matching is done by using different similarity measures such as euclidean distance. Results from different methods are evaluated and compared to present the effectiveness of this new method for feature extraction.