TubeTK/Events/2011.01.24

Danielle

Grassroots

• Went again through Romain's list of software + more of my own searching
• Unfortunately lots of software is based on matlab

• Promising:
  • CMVS (includes PMVS)
  • Bundler
  • Typical pipeline: Bundler calculates sparse 3D points from input images; input results into CMVS for denser point cloud (examples from blog)
  • Status - running into 64-bit issues (binary provided is looking for 32-bit system libraries) - can build from source (which is evidently buggy on 64-bit from my online searching), or find a 32-bit machine until we know it will work

• Could be promising, but under development:
  • libmv

• Not appropriate:
  • iModeller: Must place object on "object pattern" during acquisition
  • 3DSOM: Must place object on "3DSOM Pro mat" during acquistion
  • Code from Pierre Fillard: Simply merges stacks of 2D images into a 3D volume
  • hugin: Panoramas, not 3D reconstructions
  • Z-Anaglyph: Creates red-blue composite images for 3D viewing, not actual reconstructions
  • Gatan products: For electron micrography, looks dependent on their acquisition software

• Search terms
  • multi-view stereo software
  • structure from motion software

Sliding Organ

• Fixed test using a .vtk surface instead of a given normal vector image and weight image
• Fixed coordinate system bugs /w Slicer
• Style fixes
• Found bug in AnisotropicDiffusionTensorFunction - not taking spacing into consideration when calculating image derivatives
• Formalized strategy for dealing with corners - which was refuted by Marc =)