Noise Removal on mCT scan or rock

hi there

I have about 1300 sections created in a mCT scanner of a cylindrical cracked rock (sandstone).
(low energy beam unfortunately for such dense/hard material … but couldnt use more their scanner is for marine biology purposes not rocks)

So their results are noisy …
I aim on segmenting the crack (at the stack of 1300 pics) only as to compare it with initial state (this second stage involved corrosion of the crack) {already have successfully segmented and 3d modeled the initial state)
I need some help (ideas) for removing the noise and only have the crack selected at the following pic

I tried some techniques been proposed for other problems (in this forum and elsewhere) none had significant result…
Thanks in advance

original tiff here
scan-01146__rec0100.tif (7.2 MB)

before time i had the same sample (initial state) scanned in a more powerfuul mCt scanner where the results were far better like this one

though now i HAVE to do what i can with the noisy staff :disappointed_relieved:

I think it is better to try to get the sample scanned with higher resolution and/or better signal noise ratio rather than trying to remove the noise. See the attached image, that is a 2000x2000 pixel slice of a core sample from a carbon composite I did with the scanner on our campus at 135keV. This image is a raw export, no processing was done at all. The noise level depends much more on the detector than the X-ray energy. This material is pure carbon so the signal noise ratio will be better than on a sandstone though… But I have got good results on pure quartz samples as well.

Hi @steinr

Thanks for your time. Unfortunately there is no possibility to use another scanner. I have to deal with the noisy pics i got…

You know in some parts of the world, access in such kind of equipment is way too far than just simply "Apply for permission to use the equipment - finalize the dates - DONE "… if you get what I mean

thanks anyway

Yes I know that I am fortunate to have a good scanner available on my campus; I just book it and pay for the instrument rental. The scan I showed you costs about $1000 total in instrument cost and (my) labour per 2000^3 dataset.

Regarding noise removal, try “Despeckle” first. That is a quite gentle noise removal that will not distort the crack much. If that is too gentle, try the “Median” filter. There is also a “Median 3D” filter that will work better for stacks, which you have if you import your images as “Image Sequence”.

1 Like