A Novel Image Texture Fusion Scheme for Improving Multispectral Face Recognition

Recently, the research in multispectral face recognition has focused on developing an efficient framework that is capable of capturing and fusing the salient features from different spectral bands. However, few studies have explored the texture description to analyse the multispectral data. The local binary pattern (LBP) operator is one of the most powerful, attractive, and simple texture descriptor showing excellent results in terms of accuracy and computational complexity in many empirical studies. Motivated by the efficiency of this texture descriptor, we propose an extension of the basic LBP to evoke multispectral imagery analysis for face recognition. The proposed descriptor, called Multispectral Local Binary Pattern (MSLBP) captures the mutual relationships between the pixels in the different spectral bands. First, Gradientface is applied to address the variation in face images illumination. Then, the resulting MSLBP method provides input to PCA, LDA, and KFA classifiers for face recognition. These systems are implemented and compared with existing Local Binary Pattern face recognition systems. The obtained results show that the use of the proposed texture descriptor approach permit to achieve high recognition rates.