Research Methods Of Retinal Vessels Segmentation

Medical image processing plays an important role in the image processing area. With the development of digital image processing technology, the extraction methods of medical images have been improved greatly, which also help doctors to detect disease more accurately. Human retinal vessel is the only deeper non-traumatic micro-vascular, which can be observed directly. Retinal vessel and its pattern are usually used to diagnose retinal pathology. However, retinal vessel segmentation is very tough because of its small structure, blur boundary, as well as noise. Therefore, the establishment of computer-aided retinal image analysis system is significant for analysis of retinal diseases.On the basis of a large number of images processing methods and DRIVE image database models, we adopt different image segmentation algorithms to detect retinal vessels. The main processing is shown as follows:(1) Pre-processing of retinal images. By analyzing the RGB color system model, we compare each channel of retinal image contrast firstly. Based on the operations of digital image processing, including image enhancement techniques, image filtering, morphological operations and image segmentation, we improve retinal image contrast greatly during the pre-processing procedure.(2) Segmentation of retinal vessels. In this section, we adopt two different segmentation methods for retinal vessels. First method is also known as the matching filter of retinal vessels. By introducing the definition of the Gaussian filter matching function, we improve the contrast between target vessels and the background, and then use the threshold and edge detection methods to extract retinal vessels respectively. The second method is based on level set method to segment retinal vessels. We preprocess the retinal vessels by contrast-limited adaptive histogram equalization (CLAHE), and apply2D Gabor wavelet to enhance vessels. Then, we use the region scalable fitting energy to segment retinal vessels.Based on the theoretical analysis and experiments, we obtain the segmentation results of retinal vessels. The performance of our algorithm is measured by three important parameters:the average detection accuracy, sensitivity and specificity. The comparison result with different segmentation techniques indicates that our method can achieve desirable performance for retinal vessel segmentation.