A Novel Method for 3D Surface Mesh Segmentation

We propose a novel method for the problem of 3D surface mesh segmentation. This mesh segmentation method is based on differential geometry and geodesic distance information. In our algorithm, vertices of the same group of surface types are allowed to join a segment. The geodesic distance is estimated using shortest path between connected vertices, which can be computed from well known algorithms such as the Dijkstra algorithm. However, this is inefficient for large meshes. Therefore, in our algorithm, instead of applying the shortest path algorithm to the entire mesh, we locally apply it to mesh partitions so that we can avoid large computational costs. We tested our method with 3D surface meshes of synthetic objects and molecular structures.