The Analysis Of Character For Blood Vessel Of Ocular Fundus Images

With the development of the image processing applications, the technologies of medicalimage processing are widely used, the research of ocular fundus image processing are paidenough attention and get more findings. ocular fundus blood vessel is the onlyone which canbe direct observed, The feature of optic fundus blood vessel is more important for medicalapplied, start with the blood vessel feather extract, measurement, registration, recognitionand matching, blood vessels with diagnosis of related disease, blood vessels motion captureand three dimensional simulation, I do the following several study works on account of thefeathers of blood vessel in image of ocular fundus blood vessel.1. This paper addresses a new biometric technology--optic fundus blood vessel skeletonrecognition based on band tree. The green gray scale ocular fundus image is utilized. Theskeleton feature of optic fundus blood vessel is extracted at first. The centre of fundusblood vessel region acts as the origin to define as the root point, the branch acts as the bandtree classifying points, then form the band tree representation of The skeleton feature ofoptic fundus blood vessel. It can conveniently contrast the fundus blood vessel imagebecause of using the band tree.Matching two blood vessel images is to solve such problems as rotation, translation,zoom etc. After obtaining the matching reference point and transformation factor betweenthe two blood vessel images, rotate and translate the input blood vessel image in order todetermine whether the two blood vessel skeleton images have the band tree of samegeometry. Then according to formula, calculate feature point coordinates of the transformedinput blood vessel skeleton and vector curve orientation of the corresponding region. Andthen place the feature point set of the transformed input blood vessel skeleton on the featurepoint set of the template image, and examine the number of overlapping feature pointsbetween the two feature point sets. Since matching is not very precise, a pair of matchingpoints can not be completely superposed, and certain disparity exists as to position ororientation. This demands certain disparity tolerance. In response, a so-called threshold boxmethod is adopted in this paper. At first define a rectangular region around every featurepoint in the feature point set of template blood vessel skeleton as its corresponding thresholdbox. As long as a feature point of the transformed input blood vessel skeleton, after being superposed, falls into the very region and points to similar direction, the two points areregarded as a matching pair. Calculate the total number of matching points and present thematching result.Extracting features from two fundus blood vessel images follows several steps. At first,according to gray scale image, fix the brightest region of the window. The centre of regionacts as the origin to define as the root point. Then start with the horizontal direction anddeasil find out the first branch point of the three vector curves of blood vessel skeleton. Thisbranch point is regarded as feature point. And then calculate the matching reference point ofthe two fundus blood vessel images. The power method to register two images of bloodvessel is used to the classifying structure of band tree. The index of the classifying structureof band tree is contrasted.2. This paper set up the method which can divide the grade of diabetes andcardiovascular disease on the basis of extracted ocular fundus blood vessel feature, usemeasuring arteries and veins in the ocular fundus blood vessel image and quantitativecomparison as the gradational standard to divide diabetes and cardiovascular disease, so it ismore scientific. The green gray scale ocular fundus image is utilized and preprocessed suchas enhanced image, gradient edge detection, image division, remove noise, optic orientationand so on. It realize the way which measure the diameter of fundus arteries and veins vesselsin green grayscale ocular fundus image, according to the diameter, it can be decided whetherthe person is hypertension patients, and standardize the study of green gray scale ocularfundus image for hypertension patients.3. The dynamic tracing method based on feather of blood vessel shape are put forword.It make more beneficial exploratory for automatic positioning operation. In the lasertreatment of ocular fundus, usually, doctors perform an operation related to the focusposition in the ocular fundus blood vessel image, so it is useful whether can automaticallylocate the operation position bycomputer recognition and tracing technology.4. In the present paper, the mathematical models of visualized fundus which can beapplied to medical study are set up on the basis of the analysis of the anatomic features of theocular fundus for eyeball. The central diffuse back-projection method is used to reconstructthe curved surface of ocular fundus, and the contour image of blood vessel as well as theskeleton image is used to reconstruct the three dimensional of blood vessel. Besides, as forthe fundus system of blood vessel, the matching algorithm for blood vessel, the identificationalgorithm for branch, the algorithm for recognition for skeleton of blood vessel are put forward. The three dimensional simulation using these algorithms improve their precisionand enhance the definition of the blood vessel. As a result, this three dimensional simulationprovides the clinical application with the parameter recollection as well as the representationenvironment, and they are helpful to diagnosis, representation, laser treatment, operationpositioning.As a result, the studies made in ocular fundus image as well as blood vessel skeletonfeature are valuable and fulfilling for biometric technology, auxiliary diagnosis of diabetesand cardiovascular diseases, operation position of ocular fundus, it provide benefit method todeal with ocular fundus medical image processing...