Using the PoissonMesh Node

The PoissonMesh node uses information from a dense point cloud to generate a mesh that you can further use as a 3D object, in 3D modeling for instance. The PoissonMesh node is based on the Poisson Surface Reconstruction calculation method. The original source code and paper were created by Michael Kazhdan, Matthew Bolitho and Hugues Hoppe (see for more information).

Before using the PoissonMesh node, you need a dense point cloud with normals such as the one created by the PointCloudGenerator node. For more information on creating a dense point cloud, see Tracking a Dense Point Cloud. When you’ve got one, you’re ready to create a mesh:

1.   Connect the PoissonMesh node to a PointCloudGenerator node.
2.   A mesh is created automatically using the information from the PointCloudGenerator node and you can view it in the 3D view.
3.   If you want to, you can tell PoissonMesh to use the selected points only in the point cloud to create a mesh. To do this, check the Use Selection box in the properties panel. You can also use the Filtering box to use the PointCloudGenerator filtering to refine your mesh.
4.   You can also adjust other properties for your mesh on the PoissonMesh tab:

Depth - set the maximum depth (in integers) for the calculation tree used to create your mesh. You can enter values up to 10, but increasing this value increases the solver’s memory usage, so it’s a good idea to keep this as low as possible.

Scale - set the ratio between the 3D cube used to create the mesh and the bounding box of the footage it’s created from.

Solver Divide - set the depth at which the solver equation is used in the creation of your mesh. This can help you reduce your memory usage when generating complex meshes.

Iso Divide - set the depth at which the iso-surface extractor is used in extraction. This can help you reduce your memory usage when generating complex meshes.

Samples per Node - set the minimum number of sample points used to create your mesh. Use larger values if your original footage is very noisy.

Confidence - check to use the size of the normals as confidence information when generating the mesh. This may take longer, but gives you better results since your point cloud point creation gets double-checked.

Note:  You can only use the PoissonMesh node in NukeX as the information used to create the mesh is not stored in the node and therefore cannot be recreated for viewing in Nuke.