摘要
针对现有虹膜识别算法需要处理二维纹理信息,特征提取时间长,对噪声干扰比较敏感等问题,提出了基于Radon变换的多尺度虹膜识别算法.首先利用Radon变换将二维虹膜纹理信息转换为一维纹理信号,然后使用小波变换多尺度能量算子提取虹膜的特征信息,最后利用相异度函数进行匹配判别.实验结果表明,算法具有运算简单,快速等优点,同时算法能够较好地克服图像平移、缩放、旋转等几何攻击对识别带来的不利影响,具有较强的抗干扰能力和较高的识别率.
In order to solve some general problems in existing iris recognition algorithms such as sensitivity to noise and long calculation time for extracting 2-D feature etc, a new algorithm for iris identification of human eyes based on radon transform and multi-scale analysis was presented. First, radon transform was used to project the its texture to 1-D space, then multi-scale analysis was implemented to extract its feature. Energy values at different scales could effectively reflect the different energy distributions of iris texture and thus were recorded as a feature vector. At last, the feature vector was compared with model features through a dissimilarity function in order to classify different iris samples. Experiment results show that the algorithm is efficient while it is translation, rotation and scaling invariant as well.
基金
国家自然科学基金(60472026)资助
