Monitoring the shape changes of the optic disk is crucial for the early detection of glaucoma. This requires the optic disk to be segmented from the surrounding fundus. Previous methods required the boundary to be identified manually, for example, by specifying four control points on the boundary. This is not practical for routine clinical analysis, intra-observer and inter-observer variations are tend to be large. We develop an automatic boundary detection technique based on morphological filtering and an active contour model.