摘要
虹膜识别通常包含虹膜定位、特征提取以及编码和识别几个关键步骤。本文提出的算法从建立虹膜灰度图像的直方图入手,分析虹膜边界的灰度阈值,依靠投票机制完成虹膜的定位;然后根据人眼的生理构造特点将虹膜分区,再对经过分区处理形成的虹膜纹理特征图像用一维连续Gabor复小波提取虹膜纹理特征点的相角信息;最后对相角信息以二比特格雷码实现一次编码和循环差分二次编码,利用特征多项式计算进行比较的两个虹膜的特征差异矩值(P_iris),根据判定阈值(P0)判定,最终实现虹膜识别。
Iris recognition comprises some kernel processes as iris positioning, texture feature extraction, coding and recognition. A novel algorithm is proposed in this paper. At first, the histogram of monochrome image is constructed. The boundary of iris is determined by choosing proper threshold value after pixel value analysis. The position of iris is determined by polling-mechanism. The iris image is further divided into sub-regions of texture features, By means of complex-valued continuous Gabor wavelet transform, the phase information of iris texture features can be extracted. Such phase information is then encoded by binary-bit Gray coding followed by recurrent differential coding. After calculation using eigen-polynomial, the eigen-value difference matrices (P_iris) of two irises is compared and finally the recognition is made using threshold value (P0).
出处
《电路与系统学报》
CSCD
2003年第3期75-80,共6页
Journal of Circuits and Systems
