Previous | Next --- Slide 10 of 36
Back to Lecture Thumbnails
meranara

How are non-manifold shapes encoded/represented for graphics software? Because it seems like manifold-ness is essential to encode a shape well/easily

rmvenkat

Half-Edge Mesh is one way right. Or were you talking about something else?

siqiwan2

What if we need to encode the information inside? For example, if we need see a cross profile of human's body, in these case, only the surface seems not enough?

Max

@meranara You'd be surprised how far you can get using a collection of manifolds with boundary. Note that self-intersection is not a problem if the mesh is not attached to itself. However, if you really need non-manifold geometry, you could use sparse incidence matrices or (maybe) a polygon soup.

@siqiwan2 You probably want to use a spatial data structure such as a uniform grid, KD tree, or BVH (lecture coming soon). A volumetric analog of the halfedge mesh, though, would be a tetrahedral mesh, where you have tetrahedrons as well as faces (triangles), edges (segments), and vertices (points). Here a tet has four faces and a face glues two tets together.