Face image feature extraction method based on sparse tensor
Yuhao Liu · Journal of Jilin University · 2012
A Spare Tensor Principal Component Analysis(STPCA) for feature extraction based on Multi-linear Principal Component Analysis(MPCA) was proposed.This STPCA can obtain spare projection matrix by transforming the eigen-decomposition problem into linear regression problem.With the help of the spare projection matrix the effect of occlusion can be reduced in the process of feature extraction.Experiments on Georgia tech and AR face database show that the proposed method outperforms the MPCA method in terms of accuracy and the robustness to occlusion.