计算机工程与应用 ›› 2012, Vol. 48 ›› Issue (15): 34-38.

• 博士论坛 • 上一篇    下一篇

基于Gabor局部相对特征的掌纹识别

潘  新1,阮秋琦2,王艳霞3,金  一2   

  1. 1.内蒙古农业大学 计算机与信息工程学院,呼和浩特 010018
    2.北京交通大学 信息科学研究所,北京 100044 
    3.浙江师范大学 数理与信息学院,浙江 金华 321004
  • 出版日期:2012-05-21 发布日期:2012-05-30

Palmprint recognition using Gabor local relative features

PAN Xin1, RUAN Qiuqi2, WANG Yanxia3, JIN Yi2   

  1. 1.College of Computer and Information Engineering, Inner Mongolia Agricultural University, Huhhot 010018, China
    2.Institute of Information Science, Beijing Jiaotong University, Beijing 100044, China
    3.College of Mathematics, Physics and Information Engineering, Zhejiang Normal University, Jinhua, Zhejiang 321004, China
  • Online:2012-05-21 Published:2012-05-30

摘要: Gabor变换是掌纹识别中提取纹理特征的一个重要工具,但其性能易受图像的变化以及不均衡噪声等因素影响,因此提出了一种基于Gabor局部相对特征的掌纹识别算法。该算法对原始图像进行微尺度不变Gabor滤波;结合分形学的思想,将滤波后的图像分成大小相等的子域,每个子域又分成多个相同的子块,计算每个子块与它所在子域的相对方差,将所有子块的相对方差排列组成表征图像的特征向量进行识别。该算法将微尺度不变与局部相对特性统一,所提取的特征对各种变化有很强的鲁棒性,提高了识别精度和效率。实验使用北京交通大学BJTU_PalmprintDB证明该算法的有效性。

关键词: 掌纹识别, Gabor变换, 相对特征, 纹理分析

Abstract: Gabor transform is an important tool for texture analysis in palmprint recognition. However, it is sensitive to the variations and uneven noises. A novel method is proposed to extract Gabor local relative features for palmprint recognition. Micro-scale invariant Gabor filters are designed to convolute with the original images; and inspired by the fractals, Gabor filtered images are divided into partitions. The relative derivations of all partitions are calculated to compare the similarity between ranges and their resided domains, which composite the feature vectors for image representation. The algorithm combines micro-scale invariant and local relativity for robust features extraction, and therefore, the recognition performance is improved. Identification experiments are executed on BJTU_PalmprintDB to test the effectiveness of the proposed method.

Key words: palmprint recognition, Gabor transform, relative features, texture analysis