LGLG-WPCA: An Effective Texture-based Method for Face Recognition

This paper proposes an effective texture-based face feature extraction method which is based on Learning Gabor Log-Euclidean Gaussian, called LGLG-WPCA. LGLG-WPCA has the robust performance for adverse conditions such as varying poses, skin aging and uneven illumination. LGLG learns face features from the embedded multivariate Gaussian in Gabor wavelet domain using Whitening Principal Component Analysis (WPCA). In LGLG, we first employ Gabor wavelet to decompose the face, and then use the multivariate Gaussian distribution to fit Gabor subbands. Because the space of Gaussian is a Riemannian manifold and it is difficult to incorporate learning mechanism in the model. To address this issue, we use L$^2$EMG\cite{Li2017Local} to map the multidimensional Gaussian model to the linear space, and then use WPCA to learn facial features. Experiments show that our proposed method is an effective and promising face texture feature extraction technique.