人眼虹膜图像检测与提取方法

摘要

关于虹膜优化检测问题,由于人眼图像中存在眼睑、瞳孔等无关区域,形成的躁声信息严重影响识别精度,针对难以提取准确虹膜图像,提出了一种高效的利用面积分微分算子的直接虹膜提取方法,采用John Daugman积分微分算子进行边界检测.首先以虹膜图像灰度直方图为基础,通过灰度投影定位虹膜内边缘,然后利用积分微分算子确定虹膜左右外边界.最后,利用面积分微分算子求解最恰当的虹膜上下边界.识别过程中,利用Gabor小波提取虹膜图像特征,通过比对虹膜特征来识别虹膜.试验结果表明,改进方法可以有效地实现虹膜识别准确性.%Iris has stability, reliability and non-invasiveness. So iris can attain an higher recognition rate than other biometrics methods such as fingerprint, face and voice recognition. But it is difficult to extract iris from an eye image. We proposed a scheme of iris detection and extraction based on John Daugman' s Integro-differential operator. First, based on the original image gray histogram, the image was projected in the vertical and horizontal direction to approximately estimate the center coordinates and then get the inside edge of iris. After that, we computed the outer edge of iris using the area integero-differential operator. In recognition process, we used Gabor for extracting feature of iris and compared the Hamming distance between two iris features for recognition. The design is provably effective for one to one iris recognition through the experiment.