基于定向局部对比度的眼底图像视盘检测方法

Optic Disc Detection on Retinal Images Based on Directional Local Contrast

视盘作为眼底图像的一个重要特征,其自动检测方法在眼底病变图像分析中有着重要的作用。提出一种基于定向局部对比度滤波的方法,有效地提取眼底图像中的局部亮度区域;结合视盘区域的局部血管特征,选择定位出正确的视盘感兴趣区域;采用数学形态学方法和区域主动轮廓模型,可较准确地检测出视盘轮廓。对开放的STARE数据库上的81幅眼底图像进行测试,其中含31幅正常和50幅病变图像(含严重病理图像),用该方法正确检测出视盘73幅,准确率约为90.1%。结果表明,该方法有效地克服大块亮斑病灶对视盘检测的影响,且仅需提取粗血管,计算较为简单,说明了算法的有效性。

The optic disc （OD） is considered as one of the main features of a retinal image.OD detection is a main step in developing automated analysis systems for the retinopathy.In this paper,we proposed a method aiming to effectively extract the local brightness areas in digital retinal fundus images based on directional local contrast （DLC） filter.The right OD region of interest （ROI） was selected by considering its local vessels features.Thereupon,in the selected ROI,the OD contour was exactly detected using the mathematical morphology method and region-based active contour model technique.The evaluation for the proposed method was performed using the dataset of 81 images from the public STARE project,containing images from both normal （31） and pathological （50） subjects,even including severe pathological situations.The OD position was correctly identified in 73 out of 81 images,the accuracy rate is about 90.1％.These results demonstrate that the method effectively overcomes the influence from the large bright spot lesions on OD detection by just extracting the large vessels and less computation time.