Retinal Vasculature Extraction using Non-Subsampled Contourlet Transform and Multi-structure Element Morphology by Reconstruction

Retinal vasculature extraction is an area of utmostinterest in ophthalmology. It helps to diagnose various diseasesand also play a crucial role in treatment planning andaccomplishment.In this work, we suggest an algorithm tosegmentretinal vasculature fromretinal Fundus Images(FI) usingmulti-structure element morphology after enhancing the imageusing Normal Inverse Gaussian (NIG) model in the fuzzifiedNon-Subsampled Contourlet Transform (NSCT) domain. Sinceboth noises and weak edges produce low magnitude NSCTcoefficients, image enhancement methods amplify weak edges aswell as noises. Direct application of image boosting technique inthe NSCT domain causes over enhancement. So a novel imageenhancement method is employed by interpreting the term“contrast” as a qualitative instead of a quantitative measure of theimage. Membership values of NSCT coefficients are modifiedusing NIG model. Mathematical Morphology(MM) byMulti-structure Elements (MEs) is used to extract the edges ofimage. False vessel ridges are expunged, and the thin vessel edgesare preserved using opening by reconstruction. Connectedcomponent analysis followed by length filtering is used to filter thestill remaining false edges. In most of the available literature,low-resolution fundus image databases are used for evaluating thealgorithm. In our work, we evaluate our algorithm not onlyutilizing the DRIVE database, a low-resolution retinal image (RI)database, but also using an openly available High-ResolutionFundus (HRF) image database. Our result illustrates that theproposed method outperforms the other techniques consideredwith average accuracy (ACC) of 96.71%. In addition to ACC, wealso use F1-Score and Mathews Correlation Coefficient (MCC) toevaluate our method. The average values of the results obtainedwith the HRF image database for F1-Score and MCC are 0.8172and 0.8031, respectively, which are very much encouraging