Study On Measurement Methods Of Vascular Parameters Of Color Fundus Images

Many systemic diseases can make changes in retinal vessels, such as diameter thinning or expanding, and the changes in the arteriovenous crossing. Therefore, for the diagnosis of the diseases, it is of great significance to make a quantitative analysis of the retinal vessels.To achieve automatic measurement of vessel parameters in retinal images, this thesis proposes a method for automatic measurement of the diameters of the artery and the vein and vessel crossing angles. Firstly, on the basis of preprocessing of the retinal image, the multi-scale enhancement filter based on the Hessian matrix is used to enhance the retinal image, and the whole vessel network is segmented. Secondly, on the basis of the binary vessel and its skeleton image, the relationship between the eigenvalue and the eigenvector of the Hessian matrix is used to accurately locate the direction of the vessels, and the vessel diameters are measured by finding the vessel boundaries. After this, in order to distinguish the artery and the vein, and obtain the diameters of the artery and the vein automatically, the generalized regression neural network classifier is used to classify the artery and the vein; At last, the crossing points of the vessels are located, and K-mean clustering method is used to calculate the crossing angles of the vessels.This method is tested on the public DRIVE database and REVIEW database, which are tested to verify the performance of vessel segmentation, diameter measurement, vessel classification and crossing angle measurement. The experimental results show the effectiveness of the method.