|New MCT-based Face Recognition under Varying Lighting Conditions
Sue-Kyeong Park and Dong-Gyu Sim*
International Journal of Control, Automation, and Systems, vol. 9, no. 3, pp.542-549, 2011
Abstract : This paper presents a new face recognition algorithm that is insensitive to variations in lighting condi-tions. In the proposed algorithm, the MCT (Modified Census Transform) was embedded to extract the local facial features that are invariant under illumination changes. In this study, we also employed an appearance-based method to incorporate both local and global features. First, input facial images are transformed by the MCT and a bit string from the MCT is converted to a decimal number to generate an MCT domain image. This domain image is recognized using principle component analysis (PCA) or linear discriminate analysis (LDA). Experimental results reveal that the recognition rate of the proposed approach is better than that of conventional appearance-based algorithms by approxi-mately 20% for the Yale B database, in the case of severe variations in illumination conditions. We al-so found that the proposed algorithm yields better performance for the Yale database for various face expressions, eye-wear, and lighting conditions.
Keyword : Face detection, face recognition, feature extraction, MCT.