Parallel Heat Kernel Volume Based Local Binary Pattern on Multi-Orientation Planes for Face Representation

Appropriate representation is one of the keys to successful face recognition technologies. Actual facial appearance sometimes differs dramatically because of variations in pose, illumination, expression, and occlusion. However, existing face representation methods remain insufficiently powerful and robust. Hence, we propose a new feature extraction approach for face representation based on heat kernel volume and local binary patterns. Multi-scale heat kernel faces are created in our proposed framework. We then reformulate these multi-scale heat kernel faces as three-dimensional volume. Furthermore, we generate multi-orientation planes from the heat kernel volume, which reflects orientation co-occurrence statistics among different heat kernel faces. Finally, we apply local binary pattern (LBP) analysis on the multi-orientation planes of the heat kernel volume to capture the microstructure and macrostructure of face appearance. Hence, we obtain the heat kernel volume based local binary pattern on multi-orientation planes (HKV-LBP-MOP) descriptor. The proposed method is successfully be paralleled. We applied the method to face recognition and obtain the performance of 99.28 and 87.82% on ORL and Yale databases respectively. Experimental results on the show that the proposed algorithm significantly outperforms other well-known approaches in terms of recognition rate.