Automatic framework for extraction of red lesion using gabor filter from fundus image

Abstract

The paper proposes an automated computer vision method for detection of red lesions present in the fundus images. In case of Diabetic Retinopathy, red lesions constitute of both microaneurysms and haemorrhages. In the proposed work, the possible candidate pixels similar to red lesions with respect to intensity levels are identified and subjected to a logical subtraction operation that leads to desirable result in segmenting out the lesions only. Strategic use of Gabor filter helps in identification of blood vessels and geometrical features-based thresholding is applied to extract red lesions efficiently and make the algorithm effective. Developed algorithm is tested on a comprehensive digital fundus image database and the obtained results are encouraging with a high accuracy and has low computation cost and can be deployed for the detection of the red lesions from fundus images.