Robust Facial Recognition Based on Local Gaussian Structural Pattern

In this paper, we propose a novel local feature descriptor, Local Gaussian Structural Pattern (LGSP), for face recognition that encodes the directional information of the face’s textures (i.e., the texture’s structure), producing a more discriminant code than other state-of-the-art methods. Each micro-pattern’s structure is computed by using a derivative-Gaussian compass mask, which is more robust against challenging illumination and noisy conditions, and encoded by using the principal directions. Consequently, the compass mask helps distinguishing among similar structural patterns. Moreover, our descriptor encodes more information by using different resolutions of the compass mask. Thus to process a face, we divide it into several regions and extract the distribution of the LGSP features from them. Then, we concatenate these features into a feature vector, and we use it as face descriptor. We perform several experiments in which our descriptor showed consistent results under age, illumination, expression, and noise variations.