Combination of dual-tree complex wavelet and SVM for face recognition

Based on the attractive property such as shift invariance, good directional selectivity, limited redundancy and efficient computation of dual-tree complex wavelet transform, a novel face recognition method with combining of dual-tree complex wavelet transform and support vector machine is proposed in this paper. Firstly, it uses 2D dual-tree complex wavelet transform to decompose each face image into six band-pass sub-images that are strongly oriented at 6 different angles and two low-pass sub-images and extracts the human face features. Then principal component analysis technique is used to reduce the feature dimensions. Finally, support vector machine is used as classifier. Through the comparative experiments between the Gabor wavelet approach and the 2D dual-tree complex wavelet transform approach, the results show that the proposed approach can achieve higher recognition rate no matter what SVM kernel is used. Also, experiments show that the proposed method needs least computation time.