Towards Automatic Retinal Blood Vessels Segmentation in Retinal Images

Abstract

Nowadays, segmenting objects is desired to timely diagnose various diseases. This task is challenging as blood vessels share the same color and intensity information in retinal image area. Therefore, an accurate vessel segmentation method is required. This study presents an automated vessels segmentation algorithm. Our method initially extracts the Green channel on which the CLAHE and Gabor filter is applied. Final segmentation is achieved using Otsu’s thresholding. Meanwhile, to reduce noise from tiny vessels, median filter, Top-hat transform and other morphological operations, such as spur operation is applied in post-processing stage. The proposed algorithm yields superior accuracy results on DRIVE and STARE than several methods and consumes nearly 1 second to produce the segmented output image.