Enhanced Discriminant Local Direction Pattern Learning for Robust Palmprint Identification

Direction-based methods have been widely used in palmprint recognition methods. However, most existing palmprint direction patterns-based methods need rich prior knowledge, and usually ignores the relationships among different samples. Furthermore, how to make the extracted features more discriminative is also a dilemma to improving the recognition performance. To solve these problems, we propose to learn enhanced discriminative direction pattern in this study. We first extract the complete and stable local direction patterns, where a salient convolution average feature (EDL) is extracted from the palmprint image. Afterwards, a linear regression learning model is introduced to enhance the discriminant of EDL, such that the representation of the direction pattern can be improved. Experimental results on 4 real-world palmprint databases show that the proposed method can outperform the other state-of-the-art related methods.