基于改进型Gbvs模型的眼底图像视盘检测方法  

Optic disc detection method based of fundus image on modified Gbvs model

作　　者：吴骏[1,2] 张亚男 肖志涛[1,2] 耿磊[1,2] 张芳[1,2] 杨嵩[1,2] 张东霞[1,2] 宋舒雅 WU Jun;ZHANG Ya-nan;XIAO Zhi-tao;GENG Lei;ZHANG Fang;YANG Song;ZHANG Dong-xia;SONG Shu-ya(School of Electronics and Information Engineering,Tianjin Polytechnic University,Tianjin 300387,China;Tianjin Key Laboratory of Optoelectronic Detection Technology and System,Tianjin 300387,China)

机构地区：[1]天津工业大学电子与信息工程学院,天津300387 [2]天津工业大学天津市光电检测技术与系统重点实验室,天津300387

出　　处：《天津工业大学学报》2018年第1期54-61,共8页Journal of Tiangong University

基　　金：国家自然科学基金资助项目(61401439);天津市科技支撑计划重点项目(13ZCZDGX02100);天津市应用基础与前沿技术研究计划项目(15JCYBJC16600);高等学校博士学科点专项科研基金(20131201110001)

摘　　要：为了改善现有的视盘定位方法没有充分利用视盘视觉特征的不足,提高定位准确率和分割精度,结合视觉注意机制与相位一致性,提出一种新的视盘检测方法.首先对Gbvs模型进行改进,提取亮度、对比度和符合人类视觉感知特性的相位一致性特征,并利用改进型Gbvs模型构造显著图;接着用滑动窗口扫描总显著图,将扫描所得的显著性最高的位置视为视盘中心位置;最后消除视盘区域的血管,利用C-V模型获得视盘边界.在MESSIDOR眼底图像数据集上对本文方法进行测试.结果表明:平均定位准确率为98.83%,高于现有代表性方法,实验结果表明本文方法具有较高的定位准确率和分割精度.For resolving the problem that the visual features of optic disc(OD)were not utilized efficiently in existing methods of(OD)localization,improving the location accuracy and segementation accuracy,a new OD detection method combining visual attention mechanism and phase congruency is presented.Firstly,Gbvs model is improved by extracting three types of features including intensity,contrast and phase congruency(PC)which accords with human visual perception characteristics,and the modified Gbvs model is used to obtain saliency map.Then,the total saliency map is scanned using a sliding window,where the highest significance is scanned as the center of(OD).Finally,the main blood vessels around(OD)vicinity are removed and the boundary of(OD)is detected using C-V model.This method is tested on MESSIDOR dataset,and its average test result is 98.83%which is higher than existing representative methods.The experimental results show that the proposed method has higher location accuracy and segmentation accuracy.

关 键 词：视觉注意机制 相位一致性 视觉特征 Gbvs模型 眼底图像 视盘检测 

分 类 号：TP391.413[自动化与计算机技术—计算机应用技术] R770.4[自动化与计算机技术—计算机科学与技术]