Research Of Blood Vessel Segmentation And Optic Disc Detection Algorithm In Fundus Images Based On Pulse Coupled Neural Network

Currently,the blood vessel segmentation and optic disc detection technology based on manual method have the disadvantages of high cost,low efficiency and the detection affected by physician subjective and other problems.With the rapid development of information technology,image processing technology,pattern recognition and artificial intelligence technology,it is urgent to further study the automatic detection technology of blood vessel and optic disc in fundus images.It is very important to realize the automatic detection of blood vessel and optic disc in fundus images,which has very important social and economic benefits.In this paper,the key steps of blood vessel segmentation and optic disc detection in fundus images are studied,including the preprocessing,blood vessel segmentation and optic disc detection of fundus images.The main research results are as follows:1.A preprocessing method for fundus images is studied,for removing the disturbance in fundus images and preparing for the subsequent image processing.The main steps are as follows:(1)To remove the noise,choose adaptive median filter algorithm and adaptive smoothing filter algorithm.(2)To correct light unevenness,use geometric mean filtering algorithm and gray level global transformation algorithm.(3)To achieve vascular enhancement,apply the morphological top hat transformation algorithm.The simulation results show that the preprocessing method can better retain the information of the fundus structure and improve the contrast between the target and the background.2.A blood vessel segmentation method based on Pulse Coupled Neural Network(PCNN)is improved.The method simplifies the traditional PCNN model,and improves its single neuron excitatory link input into the sum of the neuron excitability and neighborhood inhibition link input,and improves the threshold of the time exponential decay to the gray level of the image,and improves the link strength to Laplasse energy.When the image is segmented,there is no need to set parameters manually,no specific criteria are needed to determine the best number of iterations,and one iteration process completes the segmentation,which solves the problem that the network parameters of the traditional PCNN model are numerous and difficult to choose,and the end condition of the iteration number is not well determined.The simulation experimental results of DRIVE fundus image library and the fundus image of the Affiliated Hospital of Guangxi Medical University show that the method is superior to traditional PCNN method in subjective visual effect and objective segmentation performance and operation time.3.The optic disc positioning algorithm based on template matching is implemented,the focus is to improve the Hough transform-based optic disc segmentation algorithm,crop the region of interest,substitute the edge points of the image space into the parametric equations,and compare the edge points and thresholds value of the circle.This method overcomes the shortcomings of Hough transform in detecting round time and space consumption.Through simulation experiments of fundus images of the DRIVE fundus image library and the Affiliated Hospital of Guangxi Medical University,it is shown that this method can accurately detect the optic disc,which has strong effectiveness and superiority.