A Novel Local Binary Operator Based on Discretization for Finger Vein Recognition

Chengcheng Zhao, Huimin Lu, Yupeng Li, Weiye Liu, Ruoran Gao

Abstract: It is well known that since the binary sequence and weights of the local binary operator are obtained based on local features, the local binary operator has a good description of local texture features. These algorithms have been widely used in finger vein feature extraction. However, this type of algorithm is less effective in describing the global information of the image. In this paper, we propose a new method called Discrete Binary Pattern (DBP) for obtaining binary sequences and weights. This method obtains the DBP code by calculating the mean and standard deviation in the local region of pixels and combining them efficiently, then the combined data is used as the basis to obtain the binary sequence and weight of DBP. This method overcomes the poor extraction of global image information by local binary operators. To test the effectiveness of the algorithm, we compared DBP with other representative local binary operators on three finger vein datasets, FV-USM, UTFVP, and SDUMLA. The EER of DBP achieves state-of-the-art performance in the three databases, and its AUC value is stable above 0.90 in experimental data. The accuracy of DBP is equal to the performance of LDP algorithm, but its time complexity is better than that of LDP algorithm.