Analysis And Research Of The Retinal Images

Retinal fundus image is widely used in the diagnosis of various retinal diseases:, such as glaucoma, cataracts, age-related macular degenerative diseases, etc, and even can help to diagnose hypertension, diabetes, stroke and other systemic diseases. Computer image processing and analysis techniques can improve the accuracy and efficiency of the diagnosis of the fundus images. Since the optic disc identification is the prerequisite and basis for the other retinal structure analysis, this paper target on the OD location and OD segmentation.This paper firstly describes the OD location and segmentation related technologies, then describe the pre-processing of the images, OD location and segmentation.At the stage of the pre-processing, Images are converting to grayscale according to the PCA, Then OTSU is used to determine the areas of the retina. After that, image is enhanced to reduce the effect of uneven illumination and enhance the contrast. Finally vessel structures are obtained by threshold after IUWT. Then use the pixels values around the vessels to eliminate the vessels.At the stage of OD location, this paper presents a novel locating method according to the distribution of blood vessels and OD’ shape and gray level characteristics. First, we find the horizontal position by normalized gradient direction histograms according to the fact that vertical structure of the vessels around the OD is more than the horizontal structure. Then find the vertical position by DOG around the horizontal position. Experiments in DRIONS-DB, Messidor, STARE three databases show a good locating effect.At the stage of the OD segmentation, this paper proposed a novel method which based on ray edge detection. This method dense sample the center of OD to find the pixels which get the highest grey level difference at all direction of each ray, then superimposed all the points to constitute a binary image. At last ACM is used to find the edge. Moreover, this paper ameliorates the method based on the watershed transform. After using the watershed transform to obtain respective small regions in the OD, ACM is applied to fit curve of the boundary of the small regions instead of the original circle fit method, whose precision is higher. The two method’s experiments in DRIONS-DB, Messidor two databases show a good segment effect...