Identification of the optic disk boundary in retinal images using active contours
This paper describes a technique for identifying the boundary of the optic disk in digital images of the retina, using an approach based on active contours (snakes). The optic disk is the region on the retina at which optic nerve axons enter and leave the eye. Changes in optic disk shape and area may indicate disease processes, particularly glaucoma, and accurate identification of the disk boundary may be used to quantify changes. For accurate boundary identification, some pre-processing of the image is necessary. This pre-processing minimises incorrect boundary detection due to blood vessels crossing the optic disk. Pre-processing techniques based on local minima detection and morphological filtering were developed. After pre-processing, the optic disk boundary was determined using an active contour. The contour was driven by a novel external image-derived field called the Gradient Vector Flow. This reduced the need for accurate initialisation of the contour. Results obtained by applying these techniques to a set of nine retinal images are presented.