Illumination-robust face recognition using locally directional intensity mapping and order features
Qing Zhang, Heping Chen, Hongping Fang, Shiqian Wu · 2017
Non-uniform illumination poses great challenge for a real face recognition system. To this end, a technique, which combines locally directional intensity mapping with order features, is presented to achieve illumination-robust face recognition. The idea of directional intensity mapping implies to map intensities using reliable information, while the local mapping strategy copes with the non-uniform illumination, such as shading issue. To alleviate the effects of noise and non-uniform illumination, the extended local ternary pattern (ELTP) is employed to represent intensity-invariant features of the underlying images. The recognition is performed using a Nearest Neighbor classifier in terms of Chi Square as a distance measure. Experiments on Extended Yale Database demonstrate that the proposed method can effectively solve large illumination changes in both intensity and direction. Results show that the proposed method achieves more than 92% accuracy even if the lighting direction varies from 0° to 85°.