The Research On Retinal Structure Detection

With the development of computer science, it occupies an indispensable position in various fields. What’s more, in the medical field, computer image processing technology is even more shining, which has provided a great help for clinical treatment. In China, due to the fact of the aging problem, ophthalmic diseases, such as glaucoma, cataract and diabetes complications have deeply affect the healthy of human.Retinal is known as the neurosensory of the eyeball; moreover, lots of ophthalmic diseases are occurred from it. It goes without saying that the importance of retinal. Since the rapid development of computer image processing technology, it can be used to locate the physical structures(optic disk, retinal vascular and macula) of retina automatically.Recent years, many researches about automatic locating optic disk, retinal vascular and macula have been published. Due to the impact of retinal hemorrhage, soft exudates and hard exudates, those algorithms do not take both efficiency and quality into account. New perspectives are proposed in this paper by taking full advantage of the brightness and shape to locate the center of OD(optic disk), retinal vascular and macula.There are three characteristics about OD in the retina images. First of all, the area of OD is brighter than any other place in retina images. Secondly, OD appears in roundness or ellipse. Moreover, retinal vascular comes from the center of OD and spreads all over the retinal. In this paper, we put forward a new idea of combing gray variance method with template matching to detect OD center. Candidate points are selected by gray variance method, which are used by template matching to locate OD center.Macula is an important vessel-free area near center retina. At the center of macula is the fovea. The gray level in macula is lower than the background. Based on the fact that the distance between OD center and macular is 2.5 times OD diameter, as well as the macula center is located in the centre point of artery. Our paper originally uses position information and gray information to get candidate region and least square method to perform circle fitting to detect macula center. Different from other papers, we use retinal artery to fit the circle.Compared with conventional algorithms, the method we propose can effectively localize the OD center and macula center on retina images with inadequate filed clarity and diseased conditions. Our experiment uses the four databases, namely, DRIVE,STARE,DIARETDB0 and DIARETDB1. The accuracy of the proposed method applied in the four databases is 100%, 90%, 92% and 91% respectively in locating OD center. The correctness of the proposed for macular center is 100%,86%, 91% and 89%, respectively.