Local Discriminant Direction Binary Pattern for Palmprint Representation and Recognition
Local Discriminant Direction Binary Pattern for Palmprint Representation and Recognition
复制标题
用于掌纹表示和识别的局部判别方向二进制模式
DOI：
10.1109/tcsvt.2019.2890835
复制
发表时间：
2020-02-01
影响因子：
8.4
通讯作者：
Wen, Jie
中科院分区：
文献类型：
--
作者：
Fei, Lunke;Zhang, Bob;Wen, Jie
基于方向的方法是最强大且最受欢迎的掌纹识别方法。然而，目前没有现有研究完全分析不同基于方向的方法之间的本质差异，并探索掌纹最具判别性的方向表示。在本文中，我们试图建立方向特征提取模型与方向特征可判别性之间的联系，并且我们提出一种新的指数与高斯融合模型（EGM）来表征不同方向的判别能力。EGM可以为我们提供对掌纹最优方向特征选择的新见解。此外，我们提出一种局部判别方向二值模式（LDDBP）来完整表示掌纹的方向特征。在EGM的指导下，可以利用最具判别性的方向来形成基于LDDBP的描述符，用于掌纹表示和识别。在四个广泛使用的掌纹数据库上进行的大量实验结果证明了所提出的LDDBP方法相对于最先进的基于方向的方法的优越性。
Direction-based methods are the most powerful and popular palmprint recognition methods. However, there is no existing work that completely analyzes the essential differences among different direction-based methods and explores the most discriminant direction representation of a palmprint. In this paper, we attempt to establish the connection between the direction feature extraction model and the discriminability of direction features, and we propose a novel exponential and Gaussian fusion model (EGM) to characterize the discriminative power of different directions. The EGM can provide us with a new insight into the optimal direction feature selection of palmprints. Moreover, we propose a local discriminant direction binary pattern (LDDBP) to completely represent the direction features of a palmprint. Guided by the EGM, the most discriminant directions can be exploited to form the LDDBP-based descriptor for palmprint representation and recognition. Extensive experiment results conducted on four widely used palmprint databases demonstrate the superiority of the proposed LDDBP method over the state-of-the-art direction-based methods.