Hough 变换与非线性增强结合的非理想虹膜边界定位算法

边界定位是非理想虹膜识别的关键问题之一。非理想虹膜由于经常存在纹理过强、睫毛和眼睑遮挡、虹膜巩膜对比度较差、瞳孔位置偏移等问题，这使其边界尤其是外边界定位容易出现偏差。针对上述问题，笔者提出了一种基于非线性图像增强的非理想虹膜边界定位方法。在内边界定位中，该方法首先使用混合高斯模型得到瞳孔粗略位置，然后使用弦长均衡策略搜索虹膜内边界及其中心；在外边界定位中，首先对虹膜图像进行非线性灰度变换，再利用边缘检测和改进的 Hough 变换定位虹膜外边界。实验结果表明：本算法与经典方法相比可大大提高非理想虹膜分割的准确率。

NON -IDEAL IRIS BOUNDARY LOCALIZATION ALGORITHM BASED ON COMBINATION OF HOUGH TRANSFORM AND NONLINEAR ENHANCEMENT

Iris boundary localization is one of the key issues of an iris recognition system.For non -ideal iris images,frequently -occurred strong texture,eyelashes or eyelids occlusion,low contrast between iris and sclera, and pupil deviation,etc,will lead inaccuracy localization of iris boundaries,particularly the outer one.We investigate this issue and propose the boundaries localization for non -ideal iris images based on the nonlinear image enhancement.In the process of inner localization,we firstly employ EM algorithm to segment pupil approximately,then use the string -equilibrium technique to search iris center and the inner boundary.In outer boundary localization,we transform nonlinearly the iris intensity,and adopt edge detector and improved Hough transform to find outer boundary.The experimental results depict that our algorithm improves the non -ideal iris localization accuracy compared to the classical algorithms.