Retinal vessel segmentation is an important division of automated retinal disease screening systems. The morphological variations of the retinal blood vessels correspond with the chances of cardiovascular and other related diseases. The incorrect detection of the blood vessels because of the misclassifications get reflected in the measurement results and results in the physicians advising incorrect strategies of diagnosis. In the same way, the pharmaceutical experts who prepare the drugs for diagnosis and medicines might be misdirected by this problem of misclassification. The available vessel segment graph doesn’t help in eliminating the Optic Disc (OD) boundary there by causing OD pixels misdetection, which intersects with the blood vessels. Therefore in the earlier work, the OD segmentation is carried outmaking use of discrete anisotropic filter and Particle Swarm Optimization whereas Fuzzy Neural Network (FNN) classifier has been utilized for blood vessel segmentation. But, the morphological variations are not detected accurately without an efficient feature extraction and selection process. Therefore in this work, the extraction of the geometric properties of blood vessel features are carried outmaking use of Grey Level Co-occurrence Matrix (GLCM) and the feature selection is conducted employing Mutual Information and Naive Bayesian Classifier. In this work, the noise elimination is performed applying Modified Kalman filter with image enhancement making use of the Hybrid PCA technique. The OD segmentation process is improved by using the Discrete Anisotropic Filter and Bee colony algorithm. At last, the classification of the true blood vessels is done by making use of SVM classifier. Therefore the retinal blood vessels are accurately classified considering the morphological changes that can be noticed from the experimental results. Therefore this technique yields an effective platform for suitable medicinal preparation and a precise diagnosis of retinal diseases.