An optimal approach to detect retinal diseases by performing segmentation of retinal blood vessels using image processing

Changes in the retinal vasculature of the fundus image help in identifying retinal diseases. This is done by segmenting the retinal blood vessels. Computational approaches are preferred over traditional approaches for the segmentation of vessels as it is a time-consuming process. There are various techniques involved in the proposed Contrast Enhancement using Histogram Equalization algorithm such as image pre-processing and post-processing techniques, supervised and unsupervised learning techniques. Each stage is responsible for performing a series of actions. As the images are pre-processed, a feature vector is formed to which Principal Component Analysis is applied. The output is then subjected to k-means clustering to group the pixels obtained as vessel clusters or non-vessel clusters. The vessel clusters are not processed further, while the non-vessel clusters are subjected to ensemble classification which makes use of a decision tree along with bagging. The segmented image thus obtained is the combined result of clustering and ensemble classification technique. This segmented image thus obtained is then subjected to post-processing using morphological techniques. The images are then validated which shows that compared to the existing techniques, the proposed model for blood vessel segmentation shows 95% accuracy.