结合SLIC超像素和DBSCAN聚类的眼底图像硬性渗出检测方法

为自动检测出眼底图像中的硬性渗出,结合简单线性迭代聚类(SLIC)超像素分割算法和基于密度的聚类算法(DBSCAN),提出一种对眼底图像硬性渗出的检测方法.首先,采用SLIC超像素分割算法对彩色眼底图像进行过分割;然后,采用DBSCAN对上述分割得到的超像素进行聚类,形成簇;最后,分割出目标图像,并选用标准糖尿病视网膜病变数据库(DIARETDB0和DIARETDB1)的眼底图像验证上述组合算法的可行性.实验结果表明:算法能够快速、可靠地检测出眼底图像中的硬性渗出,具有可直接对彩色图像进行分割、特征提取的特点.

Fundus Image Hard Exudates Detection Based on SLIC Superpixels and DBSCAN Clustering

In order to detect the hard exudates in fundus images automatically, this paper presented a hard exudates detection method which combines simple linear iterative clustering(SLIC)superpixels and DBSCAN clustering algorithm to detect the Harde exudates. Firstly, an over-segmentation image was formed by algorithm of the SLIC superpixels. Next the superpixels obtained were processed using the DBSCAN method, so that the final segmentation could be generated from the clusters of superpixels. The fundus image of the standard Diabetic Retinopathy datasets of DIARETDB0 and DIARETDB1 were chosen to verify the feasibility of the method proposed. The experimental results showed that the algorithms can detect exudates effectively and reliably. Moreover, the method can be directly applied to color image segmentation and feature extraction.