Investigating image enhancement methods for better classification of retinal blood vessels into arteries and veins

Developing an automatic tool for classification of retinal blood vessels into arteries and veins has gone under special attention recently due to its importance in early diagnosis of several diseases namely, diabetes, hypertension and stroke. Indeed such pathologies make alternations in artery or vein vessel tree leading to an abnormal arteriolar-to-venular width ratio (AVR). To measure AVR, arteries and veins must be carefully separated. For this purpose, a few methods have been proposed in the literature most of which are based on feature extraction. However, different factors such as non-uniformity of lightness during the image acquisition process degrade the quality of retinal images which in turn affect the results of computer algorithms. In this paper, we investigate a number of image enhancement techniques for improving the quality of retinal images considering the specific characteristics of those images. Experimental results demonstrate the significant role of image enhancement as a preprocessing step in developing an efficient system for automatic classification of retinal blood vessels into arteries and veins.