Research Of Glaucoma Recognition Method Based On Image

The digital ocular fundus image can be used to diagnose the ocular fundus disease,such as : glaucoma,ocular fundus maculopathy and so on,and also to provide a referential basis for diabetes,hypertension and other diseases.A series of parameters in the image of patients can be obtained by the computer-aided automatic detection,which not only offers a convenience for the diagnosis of disease by doctors,but also shortens the time significantly and achieves a highly effective medical detection of diseases.Using the image processing techniques to enhance the image features,automatically identify the target area in the image,separate and extract the target image of ocular fundus,and automatically measure the morphological parameters can objectively,accurately and rapidly provide a reliable basis for clinical diagnosis.The parameters of optic disk and optic cup are important structural features of ocular fundus for the identification of glaucoma.The obtainment of cup-disk ratio of patients by computer-aided automatic detection can facilitate a convenient screening of glaucoma.This paper proposes a complete set of screening methods.Firstly the pretreatment such as channel extraction and image enhancement was performed,then by the use of threshold segmentation algorithm,edge detection algorithm,Hough Transform algorithm and watershed segmentation algorithm,a treatment was performed,the optical disk and cup had been segmented,and domain area of optical disk and cup had been automatically detected,finally the cup-disk ratio was obtained.Compared with the traditional algorithms,threshold segmentation algorithm is more convenient and more accurate in calculation.Then,this paper improved the traditional region growing algorithm in order to automatically detect and segment the optical cup.In view of the fact that,the traditional region growing algorithm is flawed by inaccuracy in selecting the seed point,poor adaptivity and incorrect segmentation result,a calculation of the geometric center of region of interest in combination with the central brightness was taken as the standard for seed point selection,this paper used the 5 by 5 template to perform an average filtering to the ocular fundus image,and proposed the valley difference norm and8-neighborhood linkup criterion by which a seed merge was performed to the ocular fundus image,finally the optical cup was precisely segmented.By the use of this method,this paper has detected 15 ocular fundus images of glaucoma and 15 ocular fundus images of healthy eye from the database of high-resolution fundus image one by one,the result of experiment showed that,this algorithm could rapidly and effectively detect the optical cup in the fundus images automatically and segment it accurately,this algorithm is stable and reliable and has a good accuracy,specificity and sensitivity as compared to the traditional algorithm.