This filter is a process object class that implements a step in the N-d watershed segmentation algorithm. It produces an ordered list (SegmentTree) of binary merges between segments at increasing saliency levels. This filter is used in conjunction with WatershedSegmenter and WatershedRelabeler to process an initial segmentation into a final relabeled volume. See itk::WatershedImageFilter for an overview of watershed segmentation.
- There are two inputs to this filter, the second is optional. (1) A table of segments (SegmentTable) derived from a labeled image. (2) Optionally, a table of predetermined equivalencies among segments (EquivalencyTable), necessary for streaming applications where segments may be joined across streamed chunk boundaries. The flag Merge must be set to true to enable this functionality.
- The output of this filter is a list of binary merges of segments at increasing saliency. This is the data structure itk::watershed::WatershedSegmentTree referred to as a ``merge tree'' in the itk::WatershedImageFilter documentation.
- There are two parameters to this filter described below.
- FloodLevel is specified as a percentage (0.0 - 1.0) of the maximum possible saliency value in the initial image from which the segment table was derived. A value of 0.0 calculates no merges. A value of 1.0 calculates all of the potential merges that can occur as the FloodLevel is increased to the maximum saliency value. Typically, there is no need to calculate merges past about 40% of the total depth. Because this is the most computationally intensive piece of the watershed segmentation algorithm, it is a good idea to tune this parameter as low as possible on larger volumes.
- Merge is a boolean flag indicating whether or not to pre-merge the segments marked as equivalent in the EquivalencyTable. This is only useful for streaming applications and is turned off by default. (TRUE == merge, FALSE == do not merge).
- See also:
The documentation for this class was generated from the following file:
Generated at Thu Nov 6 01:07:40 2008 for ITK by
1.5.1 written by Dimitri van Heesch,