Dental computer-aided design (CAD) systems have been intensively introduced to digital dentistry in recent years. As basic digital models, volumetric computed tomography (CT) images or optical surface scan data are used in most dental fields. In many fields, including orthodontics, complete teeth models are required for the diagnosis, planning and treatment purposes. In this research, we introduce a novel modeling approach combining dental CT images and an optically scanned surface to create complete individual teeth models. First, to classify crown and root regions for each set of data, corresponding pairs between two different data are determined based on their spatial relationship. The pairs are used to define the co-segmentation energy by introducing similarity and dissimilarity terms for each corresponding pair. Efficient global optimization can be performed by formulating a graph-cut problem to find the segmentation result that minimizes the energy. After classifying crown and root regions for each data set, complete individual teeth are obtained by merging the two different data sets. The advancing front method was successfully applied for merging purposes by considering the signed distance from the crown boundary of the surface mesh to the root surface of the CT. The teeth models which have detailed geometries obtained from the optically scanned surface and interstice regions recovered from volumetric data can be obtained using the proposed method. In addition, the suggested merging approach makes it possible to obtain complete teeth models from incomplete CT data with metal artifacts.