Feature Level Fusion of WLBP and HOG for Hand Dorsal Vein Recognition

In this paper, a new approach is proposed to extract features from the dorsal hand vein pattern. The modified Weber Local Binary Pattern (WLBP) is feature descriptor extracted, which effectively combines the advantages of WLD and LBP. WLBP feature vector consists of two components: Differential Excitation and LBP. The Differential excitation component derived based on Weber's law, which extracts the local salient patterns. LBP is highly discriminative, computationally efficient, and extracts the local micro-patterns. By computing the two components, we obtain two images: differential excitation image and LBP image, from which a 2D histogram for WLBP is constructed. Histogram of Oriented Gradients (HOG) feature is also extracted from same image. The proposed method is fuses WLBP feature and HOG feature and assessment are done for the hand dorsal vein recognition. Experimental results show that fusion of WLBP and HOG features performs better than WLBP. Experiments are piloted on NCUT database, which shows that proposed fusion of WLBP and HOG is more effective and powerful texture descriptor. Keywords—Biometrics, dorsal vein, differential excitation, Region Of Interest, uniform patterns, Weber’s law, Histogram of Oriented Gradients.