Pré-processamento de imagens de fundo do olho: correção de iluminação e de contraste para a identificação automática de rede de vasos

Abstract

Diseases that affect the human eye may cause partial or complete loss of vision. The appearance of blood vessels and arteries present at the ocular fundus may assist in diagnosis of diseases such as diabetic retinopathy, hypertension and arteriosclerosis. Retinal images may be obtained from different ways and may have different characteristics and defects for some error in its acquisition, such as lighting, contrast and presence or noise. Such problems induces the practice of preprocessing of the images in order to produce a desirable quality and thus make the segmentation of blood vessel accurately. Therefore, this work presents a preprocessing algorithm that aims to create an image with good quality for segmentation of blood vessels. The preprocessing algorithm was evaluates on two bases freely found on the internet (DRIVE and STARE) and their results were evaluates using a vessel segmentation algorithm. The results obtained using the proposed algorithm before the segmentation method have an average improvement of 2,65% to in sensitivity and thus it is showed to be quite efficient.