Comparison of two different types of morphological method for feature extraction of retinal vessels in colour fundus images

摘要

Retinal vessel extraction is used to provide the information about the pathology characteristic of retinal related diseases, such as diabetic retinopathy (DR). According to WHO data in 2002, DR placed at the fifth rank disease causing of blindness after cataract, glaucoma, macular degeneration and corneal opacities. In previous research works, several methods have been proposed for feature extraction of retinal vessels by using morphological methods. This paper compares two different types of morphological method. The methods are morphological opening and morphological bottom hat transformation. These methods are applied on colour fundus images from DRIVE and STARE databases. The performance is evaluated by accuracy, sensitivity and specificity. Results show that morphological bottom hat transformation performs better than morphological opening.