Algorithm Study On Fundus Vessel Segmentation Based On The Combination Of PCNN And Morphological Matching Enhancement

Retinal blood vessels are the only inner tiny ones that can be directly observed by nonwounded way in human body.Changes of retinal vessel morphological characteristics can directly reflect the influence of various diseases(such as glaucoma,cataract,diabetes,hypertension,etc.)on the retinal vascular network structure.Medical personnel can thus diagnose various diseases.Traditionally,the method of manually dividing the retinal image blood vessels in the fundus image by medical personnel is not only cumbersome,and the workload is large,but also relies too much on the subjective judgment of medical personnel,which is likely to cause misidentification of blood vessels.It is difficult to meet the needs of modern medicine.It is imperative to find a set of high-precision retinal blood vessel automatic analysis techniques.In the past few decades,the automatic extraction technology of retinal blood vessels has achieved certain research results.However,the variable structure of the retinal blood vessel is complicated.And due to the limitations of existing image acquisition techniques,some tiny blood vessels in some areas of the image tend to have low contrast.The accuracy of the extraction of blood vessels is not high,and at the same time,some blood vessels have breakpoints.Aiming at these problems existing in existing algorithms,this paper proposes an algorithm combining PCNN and morphological matching enhancement technology.The algorithm is mainly divided into two parts: image preprocessing and image segmentation.It aims to remove background noise,enhance blood vessel contrast,and achieve high precision automatic extraction of retinal blood vessels.The specific research is as follows: 1)Fundus image preprocessing based on morphological matching enhancement.Firstly,the original image is grayscale transformed,then the grayscale image is processed by CLAHE,then the enhanced image is Gaussian filtered and Gabor filtered respectively.Finally,the filtered result is performed according to the ratio of Gabor filtering and Gaussian filtering 4:6.Fusion.Pretreatment can reduce the effects of uneven illumination and low blood vessel contrast.2)Retinal vessel segmentation algorithm design based on improved RG-PCNN.Combine the PCNN model,the fast connection mechanism with the seed region growth.According to the actual application scenario,the algorithm increases the ratio of the number of pixels at the edge of the blood vessel to the total number of blood vessel pixels is less than the set threshold and the area ratio of the blood vessel to the whole image is greater than the set threshold.The seed growing area,on the other hand,prevents excessive blood vessel growth.3)Implementation of retinal vascular segmentation system.A set of retinal blood vessel automatic extraction system was built under the MATLAB platform,which is convenient for understanding the effects of preprocessing and segmentation algorithms.This thesis combines the morphological matching enhancement technique with the improved RG-PCNN segmentation model,a retinal vessel segmentation method which is suitable for fundus images is proposed.The algorithm performs well on the DRIVE and STARE fundus database,a lot of experiments have been done with the current mainstream algorithms.The experimental results show that the algorithm can effectively reduce the number of vascular breakpoints and improve the segmentation accuracy of micro-vessels.This work has promising application value.