local thickness calculation failure

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

local thickness calculation failure

Anovitz, Lawrence {Larry} M.
Hi all
    I have been trying to calculate local thicknesses for a series of neutron CT images of acid-reaction wormhole formation
In  limestones.  I’m on a Mac Pro with 64 Gb of memory and 12 cores. I’m running FIJI with iImageJ 1.52i and Java 1.8.0_172 (64 bit).

The first dataset I tried worked fine.  It was a 1360 x 1360 x 2293 pixel stack,
and used at the maximum about 45 GB of memory.  The second one, however, has been a problem.  It’s a bit bigger, 1400 x 1400 x 2307
Pixels and I kept running out of memory trying to run the complete calculation so I used the individual steps provided under the
Analysis menu.  The first step, Geometry to Distance Map, ran fine.   The second step, Distance Map to distance ridge ran out of memory
Until I imported the image stack as a TIFF virtual stack.  Then it finished, or at least gave no errors.

Where I’m stuck is the third step -  Distance Ridge to Local Thickness. It does NOT run out of memory.  Both the monitor memory utility and
Activity monitor agree.  Initially it was using about 24 GB. Then I trimmed the image (the _EDT_DR result) down to 1150 x 1150 x 2307 pixels (11.4 GB)
And running with this only uses about 14 GB.  In both cases, however, it gets to slice 24 (occasionally 23) and just stops. The program itself does not crash.
I can quit cleanly or even try things like looking for updates, and those work fine.  The calculation just seems to stop at this point with no errors, and no output.

Please let me know if anyone has a solution.  I will be happy to send the original files as test cases if you would like (these are the same experiments on which
I was trying to do the 3D fractal calculations with BoneJ that also failed, btw. Michael Doube put that under “Fractal Dimension fails with AIOOB exception
on large data #163“ on the BoneJ2 list)

Many thanks

--Larry

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html