ABSTRACT
This study presents an optimized multi-level thresholding technique for the retinal vessel segmentation in fundus images. The proposed methodology involves pre-processing steps such as illumination compensation and adaptive histogram equalization to enhance vessel visibility. The segmentation is performed using a Tsallis-based multi-level thresholding algorithm, and the results are evaluated using ground truth images. Performance metrics including sensitivity, specificity, and accuracy are calculated, demonstrating the effectiveness of the proposed method in detecting blood vessels accurately. The average values of specificity, sensitivity and are found to be higher compared to an existing method. The proposed technique also shows promising results in differentiating normal and abnormal retinal images.
