Face recognition using wavelet transform and non-negative matrix factorization

This paper demonstrates a novel subspace projection technique via Non-Negative Matrix Factorization (NMF) to represent human facial image in low frequency subband, which is able to realize through the wavelet transform. Wavelet transform (WT), is used to reduce the noise and produce a representation in the low frequency domain, and hence making the facial images insensitive to facial expression and small occlusion. After wavelet decomposition, NMF is performed to produce region or part-based representations of the images. Non-negativity is a useful constraint to generate expressiveness in the reconstruction of faces. The simulation results on Essex and ORL database show that the hybrid of NMF and the best wavelet filter will yield better verification rate and shorter training time. The optimum results of 98.5% and 95.5% are obtained from Essex and ORL Database, respectively. These results are compared with our baseline method, Principal Component Analysis (PCA).