基于小波过零点位置不等长编码匹配的虹膜识别

Iris recognition based on zero-crossing position of wavelet transform and variable-length code matching

传统的Boles小波过零点方法在进行特征匹配时需将不等长编码化为等长,由于造成编码不等长的原因十分复杂,将不等长编码强行化为等长编码十分困难,从而影响识别效果。针对此问题,提出了一种基于小波过零点位置不等长编码匹配的虹膜识别方法。首先,在小波变换第4～8分解级的每级分别计算出小波过零点的位置,并以此作为编码;然后,使用一种移位匹配方法直接对不等长编码求取距离,即将待匹配的两样本数据每进行一次相对移位就计算一次距离;最后,取最小距离作为两样本距离。在包含800张图片的虹膜图库上进行了实验,识别率可达到97.70%,与Boles的小波过零点方法相比识别率提高了5.16%,一次完整的特征提取及匹配时间<0.2s,能够很好地满足虹膜识别系统对实时应用的要求。

Changing a variable-length code into an equal-length code is required in the feature match process of a conventional algorithm based on the zero crossing of the wavelet transform proposed by Boles. The factor for the variability of the code length is complex, so it is difficult to change the variable-length code into the equal-length code compulsively and to avoid the effect of feature matching on the final recognition rate. To solve the above problem, an iris recognition algorithm based on the zerocrossing position of wavelet transform and matching using a variable-length code is proposed. Firstly, The zero-crossing position of the wavelet transform at each level from the 4th to the 8th decompose level is calculated, and then shift matchingalgorithm is used to directly calculate the distance of the variable-length code. That is to say that the distance between the two samples is calculated at each relative shift. Finally the minimal distance is chosen as the distance between the two sample. Experiments are done with iris database containing 800 images, and the experimental results show that the recognition rate obtained by this method reaches 97.70%, which is about 5.16% higher than that of Boles method. For the feature extracting time and the match time is only 0.2 s,the proposed method can satisfy the system requirements of real time applications.