Retinal Vessel Segmentation Based on Difference Image and K-Means Clustering

As retinopathies continue to be major causes of visual loss and blindness worldwide, early detection and management of these diseases will help achieve significant reduction of blindness cases. However, an efficient automatic retinal vessel segmentation approach remains a challenge. This paper presents study on the combination of difference image and K-means clustering for the segmentation of retinal vessels. K-means clustering combined with difference image based on median filter addressed the segmentation of large and thinner retinal vessels as well as the reduction of false detection around the border of the optic disc. This paper also shows that K-means combined with difference image based on median filter out-performs difference image based on mean filter and difference image based on Gaussian filter while combined with K-means for retinal vessel network segmentation. The good performance of the difference image based on median filter is however attributed to the good edge-preserving property of median filter. A maximum average accuracy of 0.9556 and a maximum average sensitivity of 0.7581 were achieved on DRIVE database. While compared with the previously used techniques on DRIVE database, the proposed technique yields higher mean sensitivity and mean accuracy rates in the same range of very good specificity.