Background:

• Rendering of the crosshair including its image planes in 3-d is done by the PlaneGeometryDataVtkMapper3D.
• Image rendering in 2-d is done by ImageVtkMapper2D.
• ImageVtkMapper2D shares its textures with PlaneGeometryDataVtkMapper3D, hence:
  • The actual image rendering is done in the instances of this 2-d mapper
  • The 3-d mapper just applies these textures to 3-d planes
• Result: everything that is rendered by the 2-d image mapper is automatically shown in 3-d as well, including binary images

Issue:

• A segmentation is a LabelSetImage
• Label set images have their own 2-d mapper
• The mechanism described in the background above does not apply

Possible solutions:

• The obvious solution
  • Also share the textures of the label set image 2-d mappers with the plane geometry data 3-d mapper
  • Extend the plane geometry data 3-d mapper to also use these textures
• The "because we can (but actually couldn't)" solution
  • We can enable volume rendering for images
  • We had an option for that in the segmentation preferences
    • Didn't work reliably because of logic bugs and the actual volume rendering (nothing or black box or actual segmentation)
  • The Image interface of LabelSetImage just provides the active layer of a label set image
    • Would require to make the volume rendering aware of label set images
  • Would need to figure out a reliable volume rendering setting to show the volume as expected
  • From the perspective of H/W resource utilization probably by far the worst solution
• The only true solution
  • Implement a 3-d mapper for label set images
  • Enables full control of 3-d rendering
  • Why stop at rendering flat masks on top of the image planes if you can have 3-d surfaces in addition
  • Could be also used for segmentation previews which look a bit quirky at the moment