Supervised facial recognition based on eigenanalysis of multiresolution and independent features

In this paper, a supervised facial recognition system is presented. In the feature extraction step, a Two Dimensional Discrete Multiwavelet Transform (2D DMWT) is used to extract useful information from the face images. The 2D DMWT is followed by a Two-Dimensional Fast Independent Component Analysis (2D FastICA) and eigendecomposition to obtain discriminating and independent features. The resulting compressed features are fed into a Neural Network (NNT) based classifier for training and testing. All techniques are tested using ORL, YALE, and FERET databases. The proposed approach shows a significant improvement in the recognition rate, storage requirements, as well as computational complexity.