A Blood Vessel Segmentation Algorithm in Retinal Images Using Morphological and Spatial Features

Abstract

Extraction of blood vessels in retinal images is helpful for ophthalmologists to screen many medical disorders. The changes in the retinal vessels due to the pathologies can be easily identified by the retinal vessel segmentation. Therefore, in this paper, we propose an automatic segmentation method to extract the blood vessels from various normal and abnormal retinal images. Our proposed algorithm uses the advantages of 2D Gabor filter, contrast adaptive enhancement filter, and a powerful region growing technique called morphological reconstruction. Moreover, to increase the sensitivity measure amount, we proposed a new processing step consisting of two main part; blood vessel centerline pixels extraction and complete width of the vessel estimation. There we use modified morphological operations with different size of structuring elements and Fuzzy clustering technique. Also, we propose a new modified background estimation method to smooth the background of the retinal images to employ robust performance analysis. The final segmented image is obtained by combining the results taken from the main two parts of the algorithm. To evaluate our proposed algorithm, we applied it on two available databases, DRIVE and STARE and it proof its better performance by achieving the greatest specificity, accuracy and sensitivity in comparison with other existence methods.