In this paper, we propose a hybrid level set model with application to the segmentation of eye socket. Compared with other level set methods, the hybrid model is more robust and accurate due to the combination of region and boundary information. It solves the leaking problem of boundary based level set method. The segment results of hybrid level set model can be used in 3-D eye socket reconstruction and calculation of hydroxyapatite implant, which are useful in computer-aided surgery. Experiments have shown that hybrid level set model has very good performance in the segmentation of eye socket.