Retinal Vessel Segmentation Using Crest Lines and Morphological Operations

Retinal vessel segmentation algorithms play an important role for the detection of numerous diseases, such as diabetic retinopathy, glaucoma and many others. The aim of this paper is report an automated method for segmentation of retinal blood vessels using the crest lines and morphological operations. Initially, each image is preprocessed and the green channel is obtained. Then, we extract the vessels crest lines using a parallel thinning algorithm for grayscale images, simultaneously we perform the sum of top-hats transforms using a linear structuring element in eight directions. Both of these two techniques are performed in the green channel of the input image. The final steps consist in intersecting the two resulting images, smoothing the borders and removing spurious objects remaining. The proposed method is simple and presented robustness and high performance on the DRIVE retinal database.