Face Recognition with Single Training Image per Person Based on Wavelet Transform and Virtual Information

One of the most challenging tasks for face recognition lies in the so-called one sample per person problem. Numerous face recognition techniques will suffer serious performance drop or even fail to work in this situation. To solving it, a method based on wavelet transform and virtual information (WV-based) is proposed in this paper. First, it performs the wavelet transform on face images, then it selects the lowest frequency part with less resolution and the major information comparing to the original. Second, it does small-angle rotation on the sample image to construct virtual samples. Finally, the PCA-based classify process is done. We use ORL face database to test our method and the experimental results show its practicality and efficiency.