Automated Combination of Operations for Retinal Blood Vessel Tree Segmentation

Blood vessel segmentation from the retinal image background is deemed as a major pointer in diagnosing retinal pathologies. Theses pathologies (diseases) typically come out as an irritation in parts of the vasculature tree, altering of thickness along blood vessels, and tortuosity problem. Ignoring retinopathy for a while leads to loss of sight issues. Accordingly, we have implemented an algorithm that contributes in extracting an indicator artifact (Blood vessel) from the retinal image to give aid to the ophthalmologists, by which they diagnose retinopathogies and cure them early. In this paper, we have proposed robust, combined operations for blood vessel tree segmentation on a 2D image. In our algorithm, we applied preprocessing operation, including image filtration, non-uniform illumination correction, and color contrast enhancement, and after that, we applied combined approaches for image segmentation and classification using: two methods of texture, adaptive threshold, and morphological operators. Moreover, we introduced methods to eliminate the image boundary as well as optic disk boundary to make the vasculature individually clear and easily tracked for a disease monitoring and automated measurement. We tested our method on a number of fundus images with different views and intensities as well as non-uniform luminosity. Our algorithm provides clearer and more precise results for ophthalmologists, and automated retinal image analysis.