基于眼底结构特征的彩色眼底图像视盘定位方法

Optic Disc Localization in Color Fundus Images Based on Fundus Structure Feature

眼底图像中视盘的大小和形状等参数是判断眼底病变的重要辅助参数,视盘的检测和定位对眼科疾病的计算机辅助诊断具有重要意义。提出一种基于眼底结构特征的彩色眼底图像视盘定位方法。首先采用基于低帽运算的方法,提取眼底图像中的静脉血管;然后基于静脉血管的结构特征,采用最小二乘抛物线拟合法初步定位视盘;最后通过滑动窗口灰度扫描的方法,精确定位视盘。在4个公开的眼底图像数据库(DRIVE、DIABETED0、STARE和MESSIDOR)中,对所提出的视盘定位方法进行测试,定位准确率分别为100%、98.6%、93.8%、99.75%,验证了该方法的准确性和通用性。

The size and the shape of optic disc（ OD） in fundus images are very important for diagnosing the fundus diseases,therefore the detection and localization of OD have important significance for computer-aided diagnosis of the ophthalmology diseases. In this paper,the OD localization method in color fundus images based on the characteristics of fundus structure was proposed. Firstly,vein vessels in fundus image were extracted based on the bot-hat transformation. Then,parabolic fitting algorithm based on the least square method was applied to get the preliminary localization of OD. Finally,the accurate localization of OD was determined through window-scanning. Comparative experimental results were obtained from four publicly available datasets（ DRIVE,DIABETED0,STARE and MESSIDOR）. Testing results were 100%,98. 6%,93. 8%,99. 75%respectively,which proved the accuracy and robust of the proposed method.