Mesh Processing

It is often necessary to change the mesh. PyMesh has built-in capabilities of commonly used operations.

Collapse Short Edges

To collapse all edges shorter than or equal to tol:

>>> mesh, info = pymesh.collapse_short_edges(mesh, tol)

The function returns two things: a new mesh with all short edges removed, and some extra information:

>>> info.keys()

>>> info["num_edge_collapse"]

>>> mesh.attribute_names

>>> mesh.get_attribute("face_sources")
array([    0,     1,     2, ..., 20109, 20110, 20111])

The face_sources attribute maps each output face to the index of the source face from the input mesh.

One can even perform this function on a raw mesh:

>>> vertices, faces, info = pymesh.collapse_short_edges_raw(
...     vertices, faces, tol)

In addition to setting an absolute threshold, one can use a relative threshold based on the average edge length:

>>> mesh, __ = pymesh.collapse_short_edges(mesh, rel_threshold=0.1)

In the above example, all edges shorter than or equal to 10% of the average edge length are collapsed.

It is well known that sharp features could be lost due to edge collapse. To avoid destroying sharp features, turn on the preserve_feature flag:

>>> mesh, __ = pymesh.collapse_short_edges(mesh, tol,
...     preserve_feature=True)

One of the main applications of this method is to simplify overly triangulated meshes. As shown in the following figure, the input mesh (top) is of very high resolution near curvy regions. With pymesh.collapse_short_edges, we can create a coarse mesh (bottom) to approximate the input shape. The quality of the approximation depends heavily on the value of tol.


Split Long Edges

Another very useful but rarely implemented mesh processing operation is to split long edges. To split all edges longer than tol:

>>> mesh, info = pymesh.split_long_edges(mesh, tol)

The return values consist of the new mesh and a dummy information field for future usage:

>>> info.keys()

The returned mesh contains all the vertices from input mesh and newly inserted vertices. Certain faces may be split. Unlike standard subdivision algorithm, the algorithm only split faces that contain long edges and leaves the rest alone.

It is also possible to operate on a raw mesh:

>>> vertices, faces, info = pymesh.split_long_edges(mesh, tol)

This method is often used to eliminate long edges appearing in sliver triangles. The following figure shows its effect.



It is possible to completely remesh the input shape by calling pymesh.collapse_short_edges and pymesh.split_long_edges iteratively in an alternating fashion. The script is based on this idea. Its effects can be seen in a remesh of the Ducky The Lop Eared Bunny example:


Remove Isolated Vertices

To remove vertices that is not part of any face or voxel:

>>> mesh, info = pymesh.remove_isolated_vertices(mesh)

In addition to the output mesh, a information dictionary is returned:

>>> info.keys()
['ori_vertex_index', 'num_vertex_removed']

>>> info["ori_vertex_index"]
array([     0,      1,      2, ..., 167015, 167016, 167017])

>>> info["num_vertex_removed"]

As usual, there is a version that operates directly on the raw mesh:

>>> vertices, face, __ = pymesh.remove_isolated_vertices_raw(
...     vertices, faces)

Remove Duplicated Vertices

Sometimes, one may need to merge vertices that are coinciding or close-by measured in Euclidean distance. For example, one may need to zip the triangles together from a triangle soup. To achieve it:

>>> mesh, info = pymesh.remove_duplicated_vertices(mesh, tol)

The argument tol defines the tolerance within which vertices are considered as duplicates. In addition to the output mesh, some information is also returned:

>>> info.keys()
['num_vertex_merged', 'index_map']

>>> info["num_vertex_merged"]

>>> info["index_map"]
array([     0,      1,      2, ..., 153568, 153569, 153570])

By default, all duplicated vertices are replaced by the vertex with the smallest index. It is sometimes useful to specify some vertices to be more important than other and their coordinates should be used as the merged vertices in the output. To achieve this:

>>> weights = mesh.vertices[:, 0];
>>> mesh, info = pymesh.remove_duplicated_vertices(mesh, tol,
...     importance=weights)

In the above example, we use the X coordinates as the importance weight. When close by vertices are merged, the coordinates of the vertex with the highest X values are used.

As usual, one can operate directly on raw meshes:

>>> vertices, faces, info = pymesh.remove_duplicated_vertices_raw(
...     vertices, faces, tol)

Remove Duplicated Faces

It is also useful to remove duplicated faces:

>>> mesh, info = pymesh.remove_duplicated_faces(mesh)

The resulting mesh and some information is returned:

>>> info.keys()

>>> info["ori_face_index"]
array([    0,     1,     2, ..., 54891, 54892, 54893])

The field ori_face_index provides the source vertex index for each output vertex.

To operate on raw meshes:

>>> vertices, faces, info = pymesh.remove_duplicated_faces(
...     vertices, faces)

Subdividing Mesh

PyMesh supports both simple and loop subdivision of a given triangular mesh:

>>> mesh = pymesh.subdivide(mesh, order=2, method="loop")
>>> mesh.get_attribute("ori_face_index")
array([ 0.,  0.,  0., ..., 95., 95., 95.])

Here are some examples of different subdivision methods: