论文部分内容阅读
眼底出血点是糖尿病视网膜病变的早期症状,准确检测眼底图像中的出血点,对于构建糖尿病视网膜病变的自动筛查系统具有重要意义,本研究提出了一种基于k均值聚类和自适应模板匹配的出血点检测方法。首先利用HSV空间亮度校正以及对比度受限自适应直方图均衡化方法对眼底图像进行预处理,然后使用k均值聚类分割出候选目标,最后利用自适应归一化互相关模板匹配与支持向量机(SVM)分类器对候选目标进行筛选,从而得到真正的出血区。采用DIARETDB数据库的219幅眼底图像进行实验,本方法在图像水平的灵敏度为100%,特异性为80%,准确率为92.4%,在病灶水平的灵敏度为89%,阳性预测值为87.3%。结果表明本方法能够实现眼底图像中出血点的自动检测。
Fundus bleeding is an early symptom of diabetic retinopathy. Accurate detection of bleeding spots in the fundus images is of great importance for the construction of an automatic screening system for diabetic retinopathy. In this study, a k-means clustering and adaptive template matching Bleeding point detection method. First, HSV spatial brightness correction and contrast-limited adaptive histogram equalization are used to preprocess the fundus images, then the candidate images are segmented using k-means clustering. Finally, adaptive normalized cross-correlation template matching and support vector machine (SVM) classifier to filter the candidate target to get a real hemorrhage area. The 219 fundus images from DIARETDB database were used for experiments. The sensitivity, specificity and accuracy of this method were 100%, 80%, 92.4%, 89% and 87.3% respectively at the lesion level. The results show that this method can achieve automatic detection of bleeding in the fundus images.