This article combines structural and statistical features for face recognition.

Statistical features are introduced by applying the Zernike moments on Gabor filters.

The HOG descriptor is used to extract local statistical features.

The proposed method outperforms other methods on ORL, Yale, AR and LFW datasets.