Face recognition based on wavelet transform and adaptive local binary pattern
Local Binary Pattern (LBP) is a very efficient local descriptor for describing image texture. In this paper, we propose a novel face recognition technique based on wavelet transform and the least square estimator to enhance the classical LBP. First, Wavelet transform is used to decompose a given image into four kinds of frequency images from which the features of that image can be extracted. Then, the least square estimation of local difference between each image pixel and its neighborhoods is used to build the adaptive LBP. Finally, the classification accuracy is computed using a nearest neighbor classifier with Chi-square as a dissimilarity measure. Experiments conducted on three face image datasets (ORL dataset and two avatar face image datasets); show that the proposed technique performs better than traditional methods (single scale) LBP and PCA, Wavelet Local Binary Pattern (WLBP) and Adaptive Local Binary Pattern (ALBP) in terms of accuracy.