A comprehensive analysis of morphological process dependent retinal blood vessel segmentation

The retinal vasculature is the source of nourishment for the retina through the flow of blood. Any disruption in this blood flow results in the deterioration of the working of the retina. Various techniques have been adopted to detect these disruptions by way of extraction of the vasculature structure. In this research work, an attempt has been made to implement a blood vessel segmentation method based on adaptive contrast enhancement for noise cancellation and morphological process for the extraction of features. The pre-processing also reduces the uneven illumination problem. The background noise pixels are removed through a post processing step to achieve well identified retinal blood vessels. The proposed segmentation method is evaluated on the available public database: DRIVE, which is commonly used. The higher specificity of 98% and lower FPR of about 2% based on the proposed algorithm leads to an improved detection of blood vessels with an accuracy of about 95%.