For computer aided Glaucoma diagnostics it is essential to robustly and automatically detect and segment the main regions, e.g. the papilla (optic nerve head), in a fundus image. In this paper an effective method for automatic papilla segmentation based on the C-V model and a shape restraint is proposed. The method is a combination between the C-V model using level sets and the elliptic shape restraint for papilla segmentation. The combination of the level set framework with a shape restraint ensures that the evolving curve stays an ellipse. Experiments verify that the method shows a good performance in detecting the papilla shapes and computing the shape feature parameters within a broad variety of fundus images. The experiment results also show that the method is robust to noise and object deformity.