Internal vs. external pore space - 3D object

Hi all, I was wondering if anyone knew a method of distinguishing/defining internal pore space from the surrounding medium/open space around an 3D clast/grain? The dataset I am working with is a 3D X-ray CT scan of a highly porous and irregularly shaped object. I have a feeling it may depend on how I define the ‘boundaries’ of my object but cant seem to find anything… can anyone help?

The image below is a slice taken from my z-stack in which the object I am interest in extracted the pore network from can be seen in white. Internal pore space and the surrounding free/open space is obviously black.


Original tif image: download

Thanks,
John

Good day John,

it would help to see an original gray-level slice.
The slice you’ve posted is binarized.

Best

Herbie

Hi Herbie,

Thanks for your response, see below the original grey-scale image. It should be the same z-slice as shown in the binary image…


Original tif image: download

John

John,

thanks for posting the original slice.

I doubt that a reasonable analysis can be done based on such images.
Evidently, the material on the one hand casts pronounced X-ray shadows and on the other the image is massively over-exposed. Consequently, you won’t be able to reliably separate pores from the surround.

When comparing the binary with the original image it becomes evident that the former is highly misleading with respect to the task.

You could try to get better CT-images, perhaps with higher gray-level resolution (e.g. 16bit), but I’m far from sure that such images will allow reasonable analyses of the desired kind.

Regards

Herbie

Hi Herbie,

Thanks for your comments. The volume is just a trial dataset so I’m not too worried about the reliability of any results. I’m more interested in the development of our approach to analysing the samples in preparation for when we re-scan.

Thanks,
John

Dear John,

please be aware that I doubt not only the reliability of the analysis but that I doubt more generally the desired analysis per se, if you can’t avoid the mentioned deficiencies:

  1. Strong shadows
  2. Over exposure

Good luck

Herbie

1 Like

Hi Herbie,

I’m pretty confident that the over exposure and shadowing defects can be overcome. The high attenuation is the result of heavily stained organic material. Reducing the degree to which the organics in the samples are stained should result in better quality image data,

Thanks,
John

Assuming that it’s possible to generate raw images of suitable quality, as @anon96376101 mentions in this thread, it is possible to isolate internal pores from the outside to which they are connected.

It’s a common issue. Take a look at our paper (Carriero et al. ) where we deal with that problem wrt. canals in bone that join the ‘outside’ space. There are a couple of ways to deal with it but essentially you have to one way or the other create a logical mask that represents the total volume of porous solid and excludes the outside (by e.g. dilating a few times and eroding a few times; blurring the greyscale then thresholding…), which you then do a binary AND operation on with the pore space in another binary image. This disconnects the pores from the outside that they’re connected to.