基于能量加权的子带杂交法在掌纹识别中的应用

Palmprint Recognition Based on Energy Weight with Cross Band Fusion

在掌纹采集过程中,由于受光照噪声的影响,以及手掌的弓形常常给掌纹采集带来噪声.基于此,提出小波变换子带杂交的一种新颖掌纹识别算法.该算法综合考虑小波同层各子带及相邻层子带分解系数的噪声特点,采用基于掌纹图像空间能量加权,再由二维主元分析(Two-di mensional Principle component Analysis,2DPCA)算法降维、去相关,最终由最小距离分类器完成掌纹识别.基于香港理工大学公布的PolyU掌纹数据库的实验,此算法正确识别率达到100%.同2DPCA算法相比,提出的算法不仅正确识别率较高,识别效率也较高.

As palmprint acquisition is often influenced by illumination,and also the arch of the palmprint resulted in much noise during the course of palmprint image shot.This paper presents palmprint recognition based on energy weight with cross band fusion.Comprehensively taking into account the noisy properties of various sub-bands in single wavelet level and decomposition coefficient of the adjacent sub-band,we employ the method of energy weight based on palmprint image,then employ Two-dimensional Principle component Analysis（2DPCA） for dimension reduction and de-correlated and finally,use a nearest neighbor classifier for palmprint recognition.Experimental results on Hong Kong Poly U palmprint experiments show that correct recognition rate can reach 100% by the method.Also right recognition rate and recognition efficiency is higher than that by 2DPCA.