基于圆形保持水平集方法的虹膜分割研究  

Research on iris segmentation based on level set method with keeping circular

作　　者：范立南[1] 欧文杰[1] 孙申申[1] 赵志超[1] 

机构地区：[1]沈阳大学信息工程学院,沈阳110044

出　　处：《计算机应用研究》2014年第4期1229-1231,1235,共4页Application Research of Computers

基　　金：辽宁省自然科学基金资助项目(20102154);辽宁省教育厅科研计划项目(L2010376)

摘　　要：虹膜分割是虹膜识别系统中最重要的环节,其分割的好坏将影响虹膜识别的准确率,而虹膜识别也是最可靠的人体生物终身身份标志之一。因此,提出了基于水平集算法的虹膜分割算法。此算法是利用水平集隐式特点与圆形形状方程显式的特点相融合确保了演化曲线在演化过程中仍保持圆形,利用其思想分割内边缘。引入自适应面积项到形状约束的CV模型中用来约束外边缘。实验结果表明,尽管眼睛睁开有限、眼镜和睫毛及眼睑等遮挡以及成像设备形成图像的角度等问题,此模型仍能取得很好的分割效果。选用区域相互重叠度——DICE作为分割算法的评价指标,由实验数据可知,提出的算法对虹膜分割是有效的。Iris segmentation is one of the most important part of iris recognition system, the good or bad of its segmentation will affect the accuracy in iris recognition, and which is one of the most reliable human biological identification for life. This paper proposed an algorithm about iris segmentation based on level sets. The combination between level set implicit characteristics and shape constraint with explicit equation ensured that the evolving curve kept circular in the process of evolution to segmentation in- ner edge. In addition, it introduced the adaptive area term to CV model with shape constraint to stop outer edge. The experimen- tal results show that, although the limit of opening the eyes, glasses and eyelashes block and imaging equipment problems form the image point of view, the proposed model can still obtain a good segmentation effect. It selects the regional overlap degree DICE as segmentation evaluation index, and the experimental results show that the proposed algorithm is effective for iris segmentation.

关 键 词：虹膜分割 水平集 CV模型 形状约束 分割评价 

分 类 号：TP391.72[自动化与计算机技术—计算机应用技术]