Finger vein identification using polydirectional local line binary pattern

Abstract

Local line binary pattern (LLBP), an improved version of local binary pattern (LBP), has the limitation that it can only extract horizontal and vertical line patterns, so that the effective information in an image may not be extracted and fully utilized. In this paper, we propose an extended LLBP method for finger vein identification, which is named polydirectional local line binary pattern (PLLBP). PLLBP can extract line pattern in any orientation. Hence, to extract line patterns with the most discriminative ability and effectively utilize them, we analyze the discriminative ability of LLBP histograms from different orientations. Then, a matching score-level fusion using sum rule is applied to achieve the final matching score for the last recognition. Experimental results demonstrate the proposed PLLBP method provides matching accuracy of 99.21%, which outperforms LLBP and the existing LBP-based methods.