Image Inpainting for Hemorrhage Detection in Mass Screening of Diabetic Retinopathy

Diabetic retinopathy (DR) is one of the main causes of vision loss. The severity of DR can be analyzed using human retinal images (fundus image). Affected fundus image consists of hemorrhages, microaneurysms, and exudates along with blood vessels. In order to accurately detect the level of severity of the disease, the blood vessels are inpainted using fast marching method (FMM). The technique implemented in this paper involves image enhancement using green channel image and histogram equalization followed by mask generation and inpainting. The severity of the disease can be categorized accurately by inpainting the blood vessels using FMM. The proposed technique is tested using standard test databases HRF and DRIVE. The algorithm can be effectively used for mass screening of DR. This technique is a fundamental step in designing computer-aided diagnosis system for ophthalmic disorders.