基于多尺度-局部方向轮廓综合的鲁棒虹膜定位算法

Robust Iris Location Based on Multi-Scale Local-Direction Contour Detection

针对现有虹膜外边缘定位算法鲁棒性差的不足,本文提出多尺度-局部方向轮廓综合实现鲁棒虹膜定位的思路.该算法首先利用Gabor分解提取虹膜图像的多尺度-局部方向轮廓,在此基础上选择多个近垂直局部方向的子图进行综合而增强近垂直局部方向的虹膜边缘轮廓;在已有虹膜内边缘定位基础上根据先验知识选择有效的关键虹膜外边缘区域(消除干扰),再采用标准Hough确定虹膜外边缘形状参数,最后利用Hough变换进行小范围精确定位.仿真实验结果表明本文提出的算法在速度、稳定性方面优于现有文献报道的虹膜定位算法.

To solve the problem of present iris location, a new iris location algorithm based on multi-scale local-direction contour detection is proposed.The multi-scale local-direction contour of iris image is decomposed by gabor wavelet,then the obvious vertical contour is reconstructed by selecting multiple scales and local directions coefficients. The binary image of main exterior iris contour is chose based on the location parmeters of the iris inner contour, then the coarse exterior iris parameters are obtained based on the standard Hough transform. In order to improve the accuracy of iris location, the center and radius are decided accurately in a small scope by using Hough transform again. Comparative simulation results show that the performance of speed and robust is better than the existed method.