An approach to retinal image segmentations using fuzzy clustering in combination with morphological filters

Retinal vessel appearance is an important feature for personal identification, information security and confidentiality. It also is a key indicator for many early diagnoses, such as diabetes and hypertension. In this approach, a extraction method of retinal vessels based on fuzzy clustering in combination with morphological filtering is proposed. By decomposing the green channel image into smooth and textured components, fuzzy clustering is firstly performed on the textured composite, then the morphological open operation with multiscale linear-like structure elements is applied to suppressing noise structures. Experimental results indicate that the method can automatically and effectively extract most of the vessel backbones and branches.