Image feature extraction based on phase spectrum

The phase information is important for image feature extraction. The proposed approach can be used to extract the illumination-free image feature based on phase-only reconstruction and phase congruency. We first construct image from unit magnitude spectrum and the original phase spectrum, then we use the Gabor wavelet is to calculate the phase congruency. The experiment results show that this approach can extract the illumination-free image feature effectively. A new approach of face recognition, EigenPC, is proposed base on the illumination-free feature extraction.