A novel face description by local multi-channel Gabor histogram sequence binary pattern

Abstract

A novel method for face description by local multi-channel Gabor histogram sequence binary pattern (M-LGHSBP) is proposed. The motivation for the M-LGHSBP model is to find more rich and canonical texture measurement and deal with the high dimension problem of the local Gabor feature vector. Firstly, the normalized face image is sampled and blocked. Secondly, the blocked image is filtered by multi-orientation Gabor filters with multi-scale, which acquires rich and canonical texture measurement. Thirdly, the multi-degree LBP is used to solve the high dimension problem of local Gabor feature and its output can express both local and global features. Finally, ICA and RBF are adopted to extract feature and class. Experimental results on ORL and YEL face database show that the proposed algorithm, which achieves recognition accuracy of above 98%, is more effective than the well known face recognition algorithms, including PCA, ICA, Gabor, Local Gabor, LBP and Gabor-ICA.