Comparing 2D matched filter response and Gabor filter methods for vessel segmentation in retinal images

Analysis of retinal blood vessels is extremely important for diagnosis and treatment of many diseases such as diabetic retinopathy, hypertension, glukoma, arteriosclerosis etc. Therefore, especially vessel segmentation is one of the most critical steps for detection and treatment of these diseases. In this paper, Gabor filter and 2D matched filter response methods, which are used to enhance blood vessels, were implemented and their performances were compared. Firstly, Gabor filter and 2D matched filter are applied to fundus image in order to enhance the form of blood vessels. Then, the output image is transformed into binary for segmentation of the blood vessels by using Otsu and P-tile thresholding methods. The results show that P-tile thresholding method is more successful at binary transformation. Having been tested on 20 images obtained from STARE database, Gabor filter and 2D matched filter methods have 81.85% and 71.89% of average true positive rate respectively.