虹膜识别中噪声的检测与处理方法

Approach of noise detecting and processing in iris recognition

眼睑、睫毛会严重干扰虹膜定位的准确性,同时也会大大降低虹膜识别的性能,是虹膜定位和识别的主要干扰噪声。为了减少这些噪声对虹膜识别的影响,提高虹膜的识别率,本文采用一种计算与水平方向成不同夹角的弧线上的梯度积分的方法来定位上眼睑,采用径向对称检测下眼睑;采用灰度阈值法检测成块的睫毛,采用改进的二维Log-Gabor小波的实部滤波器来检测稀疏的睫毛。实验表明该算法能够有效去除眼睑和睫毛造成的干扰,提高了虹膜识别性能。

Eyelids and eyelashes are the major noise factors in iris location and iris recognition. They seriously affect the accuracy of iris location, and degrade the performance of iris recognition. In order to decrease the affection and increase the recognition rate, the detection algorithms for eyelids and eyelashes are presented in this paper：, which calculate the gradient integral along different angle arcs to detect the top eyelid, utilize radial symmetry to detect the bottom eyelid; adopt gray-scale threshold method to detect the block-eyelashes and use real part of improved Log-Gabor wavelet to detect the sparse-eyelashes. Experimental results show that the proposed algorithms can effectively eliminate these noise factors and imorove the performance of iris recognition.