Abstract: As hand-dorsa vein identification is non-contact,not easily polluted,and has other unique advantages,it becomes a new research and application hotspot of biometric identification methods. The focus of this paper is how to extract hand-dorsa vein image characteristics with high identification rate and robustness. This paper briefly describes the basic principle of local binary pattern(LBP) and improved LBP methods,and analyzes the disadvantages of these methods. A novel method called multi-scale block centersymmetric LBP(MB-CSLBP) is proposed. It includes not only the image's microstructures but also macrostructures,which can give more information of the image. This method is tested on a database of 2040 near-infrared hand-dorsa vein images using MB-CSLBP features and a nearest neighbor classifier. A large number of experimental results show that the proposed method offers a better recognition result of 98.21%,outperforming the original LBP and improved LBP operators,such as CS-LBP and MB-LBP.