Research Methods Of Retinal Vessels Enhancement And Segmentation

The research of retinal blood vessel image occupies the important position in the field of image processing, and it has great practical significance in clinical medicine field.It not only can reflect the high blood pressure, diabetes and blood diseases and other systemic diseases, but also a good way to help the doctor to master the patient’s situation, so it is important to analyze the characteristics of retinal blood vessels.Traditional testing methods of artificial retinal blood vessel image,although high precision can extract the whole structure of the blood vessels,but because of the long time-consuming,it takes a large effort,so it is difficult to meet the current demand for segement.Therefore, it is very important to realize the automatic extraction of retinal blood vessels in computer system.Now,although the existing retinal blood vessel extraction methods have achieved good results, the extraction of tiny blood vessels is still not satisfactory.Therefore, in view of the characteristics of retinal blood vessels, retinal blood vessel enhancement and segementation technology is in-depth stduied in this paper. In the first, proposed a method of retinal blood vessel enhancement based on the combination of contrast limited adaptive histogram equalization(CLAHE) and Gabor filtering.CLAHE method has a very significant effect on the low-contrast image processing,and Gabor filters having directional selection and frequency selection feature, the abitity to image the vessels structure of the retina and the background of undistorted preserved,will be able to combine the two great extent improved vessels contrast with the background and highlight the characteristics of tiny blood vessels.Secondly, in retinal blood vessel segmentation, an improved multi-scale single channel linear tracking algorithm is proposed.The basic idea is to start tracking a part of the seed pixels from the image, and to meet the conditions of the pixels in the tracking process to give a high degree of confidence, after the end of all pixels will be set to quantify the reliability of the vessles network.With a connected component labeling and median filtering method of combining treatment of noise, not only can remove the speckle noise can also be seen as a continuous block borders goal to eliminate noise together. Morphology theory is used in the segmentation of postprocessing, which is expanded the breadth and depth of muti-scale algorithm to some extent.In this paper, we have carried on the experiment to the DRIVE database in the MATLAB processing platform, and compared with the classical algorithm. Experimental results show that the proposed method is best to tiny blood vessels detection. Experimental data shows that the method is robust, effective and reliable.