Segmentation of retinal blood vessel structure using Birnbaum-Saunders (fatigue life) probability distribution function

摘要

Segmentation of retinal vessels is a prominent task. Retinal blood vessels contain essential information useful for computer-based diagnosis of various retinal pathologies, such as diabetes, hypertension, etc. Here we proposed a novel matched filter approach based on Birnbaum-Saunders PDF to improve the performance of the segmentation process of retinal blood vessels. The proposed method is divided into pre-processing, matched filter-based segmentation, and postprocessing modules. Pre-processing module used to improve the quality of input retinal image. After that, design the Birnbaum Saunders-based kernel for a matched filter by selecting suitable values of the different parameters with the help of exhaustive experimental analysis. Lastly, apply the post-processing module to find the final segmented retinal blood vessel. The proposed approach is tested on the DRIVE database only. The performance parameter such as average accuracy, F1-score, area under the curve (AUC) of the proposed approach is 94.61%, 0.684, and 0.9361, respectively.