Multi-scale shift local binary pattern based-descriptor for finger-knuckle-print recognition

Local Binary Pattern (LBP) has been widely used for analyzing local texture features of an image. Several new extensions of LBP based texture descriptors have been proposed, focusing on improving the robustness to noise by using different encoding or thresholding schemes where the most widely known are Median Binary Patterns (MBP), Fuzzy LBP (FLBP), Local Quantized Patterns (LQP), and Shift LBP (SLBP). LBP based descriptors are rarely applied in Finger-Knuckle-Print (FKP) recognition and especially, SLBP-based descriptors has not been reported yet. In this paper we propose using the Multi-scale Shift Binary Pattern (MSLBP) descriptor which extends the original SLBP to multi-scale to get more robust and discriminative representation of FKP features. The classification of this new proposed feature is performed by using Principle Component Analysis and Random subspace Linear Discriminant Analysis and the results suggest that they outperform other classifiers in FKP recognition. Experiments are performed using the PolyU FKP database and the results obtained have shown that the proposed FKP recognition method achieves outstanding rank-1 recognition rate up to 95% compared to the state-of-the-art FKP approaches.