自仿射拟合的虹膜纹理分割及粗分类  

Self-affine fitting based iris texture segmentation and coarse classification

作　　者：黄静[1] 苑玮琦[1] 

机构地区：[1]沈阳工业大学信息科学与工程学院,沈阳110870

出　　处：《仪器仪表学报》2015年第4期758-767,共10页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61271365)资助项目

摘　　要：提出一种应用于虹膜图像粗分类的自仿射拟合纹理分割算法,实现虹膜肠环外边界检测。该方法在多尺度下拟合图像局部到整体信息的自仿射性,有效地在小区域内描述肠环内外纹理的差异,通过分析差异的变化规律确定肠环区域。在此基础上提出利用肠环信息的虹膜图像粗分类,分别研究了肠环位置区域分布和肠环内纹理复杂度2种分类方式。在自采的1 000幅虹膜图像库下分别进行实验,分类正确率分别为98.8%和98.7%。实验结果表明,所提出方法能够有效地实现肠环区域检测,检测到的肠环信息能够应用于大样本虹膜图像粗分类。An algorithm based on self-affine fitting texture segmentation is presented, which is applied to coarse classification of iris image to implement outer boundary detection of iris intestinal loop. The self-affinity fitted information in multi-scale image from local to global can effectively describe the differences between the texture of internal and external of intestinal loop in a small scale area, and the intestinal loop region is determined by analyzing the changing patterns. On the basis of the proposed coarse classifica- tion of iris image by using the intestinal loop information, two ways of classification for regional distribution of intestinal loop loca- tion and texture complexity of internal intestinal loop are investigated respectively. 1000 iris images collected from our database are analyzed, and classification accuracy reaches 98.8% and 98.7% respectively. Experimental results show the proposed method can effectively realize the intestinal loop region detection, and the detected intestinal loop information can be applied to a large scale sample of iris image coarse classification.

关 键 词：自仿射拟合 纹理分割 粗分类 虹膜图像 肠环区域 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]