Face recognition based on Log-Gabor and uncorrelated LDE

In view of the problem of feature extraction in face recognition, a Log-Gabor wavelet and uncorrelated LDE (LgULDE) based on algorithm for face recognition is proposed in this paper. The proposed algorithm first gets the high-order statistical information by calculating the Log-Gabor wavelet representation of face images. Face images are decomposed into their spatial or frequency domains by Log-Gabor wavelet transforms. Then an uncorrelated basis function constraint is added into LDE algorithm, which is utilized to select the Log-Gabor feature vectors. Finally, the robust and discriminative feature vectors are extracted for classification. Experiments on the ORL and PIE database demonstrate the effectiveness of the new algorithm.