Retinal Vessel Segmentation: An Efficient Graph Cut Approach with Retinex and Local Phase

摘要: Our application concerns the automated detection of vessels in retinal images to improve understanding disease mechanism, diagnosis and treatment a number systemic diseases. We propose new framework for segmenting vasculatures with much improved accuracy efficiency. The proposed consists three technical components: Retinex-based image inhomogeneity correction, local phase-based vessel enhancement graph cut-based active contour segmentation. These procedures are applied following order. Underpinned by Retinex theory, correction step aims address challenges presented intensity inhomogeneities, relatively low contrast thin compared background. phase technique is employed enhance its superiority preserving edges. method used efficiency effectiveness from enhanced using filter. have demonstrated performance applying it four public datasets (3 color fundus photography 1 fluorescein angiography). Statistical analysis demonstrates that each component can provide level expected. widely unsupervised supervised methods, showing overall outperforms competitors. For example, achieved sensitivity (0:744), specificity (0:978) (0:953) DRIVE dataset very close those manual annotations obtained second observer.