Fusion of structure adaptive filtering and mathematical morphology for vessel segmentation in fundus images of infants with Retinopathy of Prematurity

Accurate visualization of retinal vasculature is essential for the diagnosis of the severity of various vascular diseases. Therefore blood vessel segmentation becomes an indispensable part of computer-based retinal image analysis systems. Retinal fundus images of premature infants are of relatively low contrast, and hence difficult to segment, when compared to adult retina images. An efficient segmentation method to extract the blood vessel network of infant retina images by the fusion of guided filter and mathematical morphology is developed and implemented in this work. It utilizes a novel post-processing technique using modified morphological closing operation which will preserve the thinner vessels in the segmented output. The segmented result can be used for the extraction of relevant features for the diagnosis of Retinopathy of Prematurity.