October 5, 2015

Federath sent me a CDM of his turb+grav+mag+jets run. This was a fits file, whose pixel values represented column density in units of g/cm². Ds9 showed the values ranged between 0 and ~2.5. Amira doesn't take fits files, but can process image formats. So I opened the fits in gimp and then exported it as a tiff file, and read this version into Amira. This apparently rescaled the data, as you can see from this histogram of voxel values:

Now all the pixel values lay between 0 and about 66. This might have to do with the formatting of the tiff file (256 possible values in this format), but isn't overly important for skelatonization. If we wanted to do more in depth stats, might want to look into reading the data into Amira another way to preserve the absolute values of the voxels.

The "auto" skeletonization method appears to take a threshold as an input. This threshold seems to 'mask' (or 'segment') the data so to only look at voxels with values above this threshold. It then does a 'centerline' analysis, a 'distance map' analysis, and finally a 'thinning' of the skeleton so that it is only 1 voxel across. To see the paper on this algorithm, it is attached to this page. It is also possible to do these 3 steps on your own in Amira, and thus gain more control over the skeletonization process.

Here are some results testing different thresholds in auto skeletonization mode, and comparing them to Federrath's figure (the paper that contains this figure is attached to this page).

I tested a couple smaller thresholds and the number of filaments seemed to converge to this value, N = 909.

There are a lot of clear filamentous structure that Amira is missing — so I tried playing in the filament editor of Amira, which instead of a threshold now let's you auto trace using a windowing function. This windowing function lets you control contrast.

This suggests that the first way of doing the tracing was using the color map to find filaments, and then letting you adjust the threshold for this. This let you control I suppose what filaments to capture — you just make them stand out in the color map. Looking at the Fig. 5 in the algorithm paper (Fouard et al 2006) seems to suggest contrast is what is being traced.