Skip to Main Content
In face recognition system age variation causes the serious problem. In this work discriminative model of face recognition is used to deal with age invariant problem. Feature extraction is done by densely sampled local image descriptors such as Scale Invariant Feature Transform (SIFT) and Multi scale Local Binary Pattern (MLBP) which gives the discriminatory information useful to detect the edge direction of the image which is robust to illumination, pose, expression and occlusion. SIFT and MLBP features used to recognize the occluded face images also. Since the extracted features are with high dimensionality new technique called Multi Feature Discriminant Analysis (MFDA) is used to reduce the feature space. MFDA is similar to LDA, where multiple features are combined with two different random sampling methods, multiple LDA based classifiers are constructed and then result from each classifier are combined by fusion rule for final decision. Another method to reduce the dimensionality of extracted features is PCA and classification is done by nearest neighbor method. Finally the results of two methods are compared. Images are taken from FG-NET database.