Since the feature extracted by the basic LBP operator are not complete and can not fully represent the local feature of face, a face recognition algorithm based on block completed local binary pattern is proposed. Firstly, the original face image is divided into small blocks from which the local difference value and central pixel grayscale value are analyzed. Extracting the historgram statistical characteristics of each block by the Su2CLBP (8,2) 、 Mu2 CLBP (8,2) and C operator. Then, the CLBP histograms of all the blocks are linked to get the CLBP feature to be used as the CLBP(8,2) face descriptor. Finally, the classification is performed using a nearest neighbor classifier with Chi square as a dissimilarity measure. Experimental results on ORL、FERET face database show that the proposed algorithm can achieve high face recognition rate up to 99.5%、92% and 98.67%, which are 2.5%、8% and 2.67% higher than the block LBP algorithm. This work demonstrates that the completed LBP feature is complete and highly discriminable and has good performance in the ORL, FERET and YALE face database.