Face Recognition Based on MB-LBP and Improved LFDA

A algorithm on face recognition based on multi-scale block local binary patterns(MB-LBP) and improved Local fisher discriminant analysis(LFDA) was proposed,which strengthens local analysis for labeled samples and global analysis for training samples with the ability of MB-LBP for local and global description.The algorithm makes use of ave-rage Euclidean distance from every sample to other samples in the same class as the parameter to overcome the limit for computing within-class scatter and preserves global structure by inosculating the total scatter of training samples in the form of parameter.Experimental results show that MB-LBP provides good base for making analysis of local and global preserving,and improved LFDA has more obvious adaptability and recognition rate than LFDA when the number of labeled samples is small.