Abstract: 3D palmprint recognition method is proposed?based on improved Contourlet transform. Through the shape index, the method maps 3D palmprint image into gray image to overcome the shortcomings of conventional mean or Gaussian curvature mapping that the characteristics of 3D palmprint image are difficult to accurately describe. Based on this, it introduces 7/5 filter to Contourlet transform, and extracts the mean and variance of each directional subband from the shape index map in the transform domain as feature information of palmprint image, which effectively utilizes superior direction feature expression ability of the Contourlet transform and eliminates the correlation included in each subband image of traditional Contourlet transform. Finally, the paper uses Euclidean distance nearest neighbor classification method to realize the classification and recognition of the test images. The experimental results show that, according to the three-dimensional
palmprint database of the Hong Kong Polytechnic University, the overall recognition rate of the proposed method improves by 2.9% than the PCA method and has obvious advantage.

Key words: palmprint recognition, shape index, Contourlet

摘要： 提出一种基于改进Contourlet变换的3D掌纹图像识别方法；该方法通过形状指数将3D掌纹图像映射成灰度图像，以克服常用的均值或高斯曲率映射难于精确描述3D掌纹特征的缺点；基于此，将7/5滤波器引入Contourlet变换，并在变换域提取形状指数映射图各方向子带的均值与方差作为掌纹图像的特征信息，从而有效利用了Contourlet变换优越的方向特征表达能力，又可有效消除传统Contourlet变换各子图像存在的相关性；最后采用欧氏距离最近邻分类法，实现了测试图像的分类识别。实验结果表明，针对香港理工大学所提供的三维掌纹数据库，该方法总体识别率较PCA方法提高了2.9%，具有明显的优势。

关键词: 掌纹识别, 形状指数, Contourlet