A novel approach for red lesions detection using superpixel multi-feature classification in color fundus images

Since red lesions have been found to be one of the earliest lesions in diabetic retinopathy (DR), automatic red lesions detection plays a critical role in diabetic retinopathy diagnosis. In this article, we develop a novel method using superpixel segmentation and multi-feature classification (SMFC). Using our proposed method, the retinal images are segmented into superpixels with the similar color and spatial location. And then, a set of features under multi-channel are proposed for each superpixel. Specifically, a novel contextual feature is developed to describe the superpixels with red lesions. Next, FDA classification algorithm is used to classify the red lesions with multi-feature for each superpixel. Additionally, post-processing is applied to remove the blood vessels and the fovea. Experiments are carried out on public DiaretDB1 database and extensive results show the effectiveness of our proposed method.