Retinal Blood Vessel Extraction Based on Adaptive Segmentation Algorithm

Abstract

The retinal fundus image contains important information associated with the visual abnormality. The blood vessel segmentation is important and mandatory for diagnosing various ocular fundus. Diabetic retinopathy affecting the retinal vascular structures is the leading and main causes of blindness. An algorithm for retinal blood vessel segmentation from retinal fundus images has been proposed in this article. The proposed retinal blood vessel algorithm is classified into four major steps. At first, the grayscale image is generated from the input RGB fundus image and applies anisotropic diffusion method to remove blur, while preserving edge. Then, the retinal image is enhanced and applies top hat transformation approach. After that, the enhanced image is divided into small sub-image and applies local property-based intensity transformation (LPBIT). Finally, the blood vessel is segmented using k-mean clustering from each sub-image. The DRIVE and the STARE data sets have used to check the performance of the proposed vessel segmentation algorithm. The experimental result has shown that the proposed algorithm achieves around 95% accuracy and 98% specificity, which is better than many states of the art methods.