Retinal vessel segmentation based on phase congruence and GLCM sum-entropy

The detection and analysis of retinal vessels in ophthalmology is of great use in the diagnosis and progression monitoring of diabetic retinopathy. Automatic Detection of the vessel network has however been challenging due to noise from uneven contrast and illumination during the retinal image acquisition process. This paper presents a robust segmentation technique that combines phase congruence and Gray level co-occurrence matrix (GLCM) sum entropy information for the detection of vessel network. While compared with the previously used techniques on DRIVE database, the proposed technique yields high mean sensitivity and mean accuracy rates in the same range of very good specificity.