Fisherface vs. Eigenface in the Dual-tree Complex Wavelet Domain

Inspired by the superior performance of the dual-tree complex wavelet transform (DT-CWT) on signal representation and feature extraction, the well-known fisherface which is based on the linear discriminant analysis (LDA) and the eigenface which is based on the principal component analysis (PCA) are investigated in the DT-CWT domain and benchmarked in this paper. For the eigenface, the probabilistic reasoning model (PRM) classifier is incorporated in the DT-CWT-based PCA scheme to improve face recognition performance. For the fisherface, the DT-CWT-based LDA approach is first time proposed in this paper to demonstrate its much superior performance. To conduct the performance evaluation for both the new fisherface and the improved eigenface methods in the DT-CWT domain, the experiments are conducted on a widely-used AT&T face image database. As shown by the experimental results, the proposed DT-CWT-based PCA with PRM method outperforms the original scheme by 3.3% based on 100 features, while the proposed DT-CWT-based LDA approach achieves the best performance regardless of the number of features used and reaches the recognition rate of 99% with 39 features.