Retinal Blood Vessel Segmentation for Diabetic Retinopathy Using Multilayered Thresholding

The most important disease that cause vessels structure to change is diabetic retinopathy. When the pancreas does not secrete enough amount of insulin diabetic retinopathy occurs. The symptoms of diabetic retinopathy include cotton wool spots, hemorrhages, hard exudates and dilated retinal veins. There are advance care of diabetic is present but still vision loss is also present. The risk of vision loss can be reduced significantly by timely diagnosis of diabetic retinopathy. For segmentation vessels play as one of the important landmark feature. Vessel segmentation algorithm can save time, patients’ vision and medical cost. The present study is focused to develop method for enhancement and segmentation of vessel. We present a method that uses Curvelet transform for vessels enhancement and multilayered thresholding technique for vessel segmentation. Curvelets are a non-adaptive technique for multi-scale object representation. Being an extension of the traditional wavelet concepts, they are becoming popular in same fields, namely in scientific computing and image processing. Curvelet transform having two main features directionality and anisotropy scaling law. These features represent the edges along curves more efficiently than the traditional wavelet. For accurate vessel segmentation morphological operation and multilayered thresholding is used. In multilayered thresholding technique, we apply different thresholds values iteratively and keep track of retinal blood vessels in successive layers. In medical imaging system, the accuracy of result is very critical; that is way we used publicly available DRIVE and STARE database.