基于微积分算子的彩色虹膜图像定位算法

针对NICE:Ⅱ中的彩色噪声虹膜图像难于精确定位问题,提出了一种基于微积分算子的彩色虹膜图像定位算法.首先选择RGB虹膜图像的R层,结合Gabor滤波器和图像梯度信息检测图像中的光斑区域;然后利用Adaboost算法检测虹膜区域,并使用抛物线形微积分算子定位上下眼睑;再利用基于微积分算子的模板,通过局部极值的逐步迭代和对虹膜边界点的聚类分析,定位虹膜外边界;最后使用同态滤波对虹膜区域进行增强处理,检测虹膜内边界.在NICE:Ⅱ彩色虹膜图像库上的实验表明,该方法的定位准确率为98.22%.

Location Algorithm of RGB Iris Image Based on Integro-Differential Operators

A location algorithm based on integro-differential operators was presented for solving the difficulty of precise segmentation in NICE:II. The R layer of RGB iris image is chosen first, and the Gabor filter and gradient information are used to detect the reflection area. Moreover, Adaboost algorithm is used to detect iris area in the whole image, then the upper and lower eyelids are detected. A center template based on integro-differential operators was adopted to search the outer boundary of iris through iteration step by step, and cluster analysis was used to remove the noise on iris boundaries. Because the inner boundary of iris is weak, the homomorphic filtering is adopted to enhance the iris area, then the inner boundary is located. The proposed algorithm was tested under NICE:II, and the accuracy of the new method is 98.22%.