Blood vessels detection and segmentation in retina using Gabor filters

Segmentation of the vessels in retina images is an essential step in earlier diagnosis of diabetic and hypertension. The application of image analysis using segmentation could assist the ophthalmologists, to detect the signs of diabetic retinopathy in the early stages. In this paper, a bank of 180 Gabor filters is used to capture high frequency information among which the maximum response for each pixel is selected. The vessels in the filtered retina images are segmented using a threshold value that passes the informative pixels and rejects the insignificant pixels. Determination of an effective threshold is of utmost importance for reliable segmentation which leads reliable vessel detection. The paper proposes a systematic way of determining the threshold value for reliable performance. The proposed approach is applied to retinal images from DRIVE retina database. The performance of the proposed vessel segmentation approach reaches to 0.95 based on the area under the receiver operating characteristic curve.