Retinal Blood Vessel Extraction From Fundus Images Using Improved Otsu Method

Related Work

Various techniques have already offered for blood vessel segmentation by taking different databases which can be generally categorized as supervised and unsupervised methods.

In supervised method, the pixels are classified either as vessel or non-vessel. Here the classifiers are accomplished with the data from manually segmented images (Dash, et al., 2017). Roychowdhury et al. (2015) introduced a three-stage novel retinal blood vessel segmentation algorithm using fundus photographs. In You et al., (2011), a new technique has been suggested by You et al., for segmentation of the retinal blood vessels using radial projection and vessel centerline detection. Marin et al. (2011), anticipated a supervised method where a 7D feature vector is take out from the input retinal images and given as input to the neural network. The classification result acquired from neural network is thresholded to categorize the pixels as vessels or non-vessels. In Soares et al., (2006), the author Soares et al. presented a technique by means of 2D Gabor wavelet and supervised classification where the process classify each image pixel either as vessel or non-vessel grounded on pixels feature vector. In Fraz et al. (2012) described a novel supervised method by employing a decision tree-based ensemble classifier to segment the retinal blood vessels. Vega et al. (2015), employed an automated segmentation method that utilizes a Lattice Neural Network with Dendritic Processing (LNNDP) to extract the retinal blood vessels from the ophthalmoscopes images. In Fraz et al, (2014), the author offered a supervised method by means of ensemble classifier of bagged decision trees to segment vessels of 9 to 10 years children of different ethnic origin. In, Guo et al. (2018) elaborated a convolutional neural network with reinforcement sample learning strategy to extract the blood vessels.