Mask/Labeling Conversion Methods

[awalter17]

Hello!

The goal of this PR is to add mechanisms for going between Labelings and MaskPredicates.

Changes this branch makes:

• Adds LabeledMaskInterval class, which wraps a MaskInterval and adds a label
• Adds methods to go from Labeling to MaskInterval
• Adds methods to go from List< MaskInterval > to ImgLabeling

Notes:

• Currently there's only LabeledMaskInterval because there's no unbounded or real-space labelings
• Labeling to MaskInterval methods just wrap LabelRegions as MaskIntervals then as LabeledMaskIntervals
• MaskInterval to ImgLabeling rasters MaskInterval to the provided ImgLabeling, which means the conversion is not fully reversible (i.e. you can go back to a MaskInterval but information has been lossed)
• MaskInterval to ImgLabeling methods will recurse through CompositeMaskPredicates, so if the root MaskInterval isn't a LabeledMaskInterval but the children are those labels will be added to the ImgLabeling. However, the labels will only be added at locations contained in the root MaskInterval
  • One method ignores MaskIntervals which are not LabeledMaskIntervals even if they're part of a CompositeMaskInterval. Meaning you could have a MaskInterval which is an "or" of one labeled and unlabeled mask, and only the labeled mask will appear in the ImgLabeling.
  • The second method allows you to set a "default label" so if there are any unlabeled pixels in a ROI this default label will be used

Please let me know if you have any questions, or if any changes are necessary!

[ctrueden]

Rebased over latest master to fix conflicts.

[ctrueden]

@awalter17 Do you need this for KNIME? I vaguely recall you mentioning work along these lines for KNIME back in September. Should we prioritize merging this?

[stelfrich]

@ctrueden If I am not mistaken, @awalter17 had invested some more time after discussions we had about KNIME's OMERO integration. @gab1one's implementation to read ROIs from OMERO is a little suboptimal and should be handled upstream..

[awalter17]

@ctrueden thanks for fixing the conflicts! It would be nice to get this merged, because I am starting forget what I did and why. However, I don't think this merge is currently blocking anyone. So I would say this merge is "nice to have", but not immediately critical.

As @stelfrich mentioned, this PR came out of discussions about KNIME's OMERO integration. As I recall, KNIME currently uses Labelings for ROIs but ImageJ-OMERO uses MaskPredicate ROIs, so the KNIME OMERO integration currently lacks ROI support. This PR would be a first step towards ROI support in the KNIME OMERO integration.