Palmprint Recognition Using Discriminant Local Line Directional Representation

Abstract

Palmprint is a new biometric feature for personal identification with a high degree of privacy and security. In this paper, we propose the palmprint feature extraction method which combines the direction-based method (Local line direction pattern) and learning-based method (two-directional two-dimensional linear discriminant analysis ((2D)2LDA)) to get the high discriminant direction based features, so-called Discriminant local line Directional Representation (DLLDR). First, the algorithm computes the LLDP features with two strategies of encoding multi-directions. Then, (2D)2LDA is applied to extract DLLDR features with higher discriminant and lower-dimensional from the LLDP matrix. The experimental results on the public databases of Hong Kong Polytechnic University demonstrate that our method is effective for palmprint recognition.