摘要：Because of the limitation for the traditional local binary pattern (LBP) operator in facial expression recognition, an effective facial expression recognition algorithm based on completed local binary pattern (CLBP) is proposed. Firstly, the method performs face segmentation and illumination normalization on original image. Secondly, CLBP operator is used to extract facial expression features from the preprocessed image. CLBP histograms could be combined by CLBP-Center (CLBP_C), CLBP-Magnitude (CLBP_M) and CLBP-Sign (CLBP_S) features in two ways, jointly or hybridly. Lastly, the nearest neighbour classifier is applied for classifying and recognizing the CLBP histograms. The experiments on JAFFE and Oulu-CASIA facial expression databases demonstrate that the proposed algorithm can achieve better performance than using the traditional LBP facial expression method. In addition, our algorithm can achieve good performance in complex illumination conditions.
