[vtk-developers] vtkDiscreteMarchingCubes

[Lorensen, William E (Research)]

Folks,
I've added a new class to Patented called vtkDiscreteMarchingCubes. We use
it to process discrete volumes, typically label maps output by segmentation
processes. It can process all the labels in a volume and creates a single
polydata. The celldata of the polydata contains the label corresponding to
the cell. 

Bill

// .SECTION Thanks
// Jim Miller at GE Research implemented the original version of this
// filter.
// This work was supported by PHS Research Grant No. 1 P41 RR13218-01
// from the National Center for Research Resources and supported by a
// grant from the DARPA, executed by the U.S. Army Medical Research
// and Materiel Command/TATRC Cooperative Agreement,
// Contract # W81XWH-04-2-0012.

// .SECTION Description vtkDiscreteMarchingCubes is a filter that
// takes as input a volume (e.g., 3D structured point set) of
// segmentation labels and generates on output one or more
// models representing the boundaries between the specified label and
// the adjacent structures.  One or more label values must be specified to
// generate the models.  The boundary positions are always defined to
// be half-way between adjacent voxels. This filter works best with
// integral scalar values.
// If ComputeScalars is on (the default), each output cell will have
// cell data that corresponds to the scalar value (segmentation label)
// of the corresponding cube. Note that this differs from vtkMarchingCubes,
// which stores the scalar value as point data. The rationale for this
// difference is that cell vertices may be shared between multiple
// cells. This also means that the resultant polydata may be
// non-manifold (cell faces may be coincident). To further process the
// polydata, users should either: 1) extract cells that have a common
// scalar value using vtkThreshold, or 2) process the data with
// filters that can handle non-manifold polydata
// (e.g. vtkWindowedSincPolyDataFilter).
// Also note, Normals and Gradients are not computed.