Weber Local Descriptor with Double Orientation Line Feature for Finger Vein Recognition

Weber Local Descriptor (WLD) is a simple and powerful local descriptor which consist of the differential excitation and the orientation. However, WLD has not been successful in finger vein recognition. In this paper, we propose a novel local descriptor for finger vein recognition, called Double Line feature Weber Local Descriptor (DLWLD), which improves two components of WLD. The differential excitation is redefined by bringing in Sobel operator, which can increase the discrimination of edge-texture. Meanwhile, the gradient orientation is replaced by double Modified Finite Radon Transform (MFRAT) orientation to obtain discriminative line feature. Moreover, cross-matching algorithm is utilized to better evaluate the similarity of features. The proposed descriptor is implemented to finger vein recognition in two public datasets. Experimental results demonstrate that DLWLD is better than WLD and other local descriptors.