Retinal Image Segmentation Based on Multiple Features Method

Vessels segmentation is one of the key issues in the study of retinal fundus image. Retinal image is the only deep microvessel that can be observed directly without trauma, and is widely used in the field of medically assisted diagnosis. It is of great clinical significance to use computer to automatically detect and segment the vascular network in retinal fundus image. In this paper, we propose a method of retinal vessels segmentation based on multiple features. There are three steps in this method. First, the energy and standard deviation in texture feature are computed and the energy feature is used to do the initial segmentation. The second step uses standard deviation texture feature to locate the optic disc region and segment the local blood vessels in the optic disc region. In the third step, Otsu algorithm is used to segment the blood vessels, and the edge pixels are extracted as supplements. In addition, morphological operation and connected domain removal operation are also used. The results prove that Multiple Features Method (MFM) is more effective than single feature. In particular, the blood vessels in the optic disc region were preserved and supplemented. The accuracy of blood vessel segmentation is improved, the blood vessel structure is more complete, and the problem of blood vessel loss in some local areas during segmentation is solved.