EXPERIMENTAL STUDY OF TRANSLATION-INVARIANT DWT FACE FEATURE ESTIMATION
Kun Ma, Xiaoou Tang · International Journal of Wavelets Multiresolution and Information Processing · 2004
In this paper, we conduct a series of experiments to evaluate the translation invariant property of a set of discrete wavelet features in a face graph. Using local-area power spectrum estimation based on discrete wavelet transform, we compute a feature vector that possesses both efficient space-frequency structure and translation invariant properties.