Segmenting and counting cracks amongst large porosities

Hello all,

Attached is a z-projection of 200 slices of a volume which I have been trying to extract the cracks from. This is carbon composite material and using z-projection allows me to enhance very fine cracks and then attempt to segment them. Unfortunately, I’ve not yet found a solution.

CRACK_COUNT_SC.tif (3.8 MB)

Attached is an image with the cracks highlighted that I would like to extract. The idea is to create an automated process as there are multitudes of volumes that I need to look at.

I have tried ridge detection, but the noise remains an issue. Also tried local thickness plugin, but not managing to remove the noise significantly. Hit a dead end, and therefore asking for help :slight_smile:

Please feel free to crop the image if necessary (to something like below) -

Thank you in advance.
Kind regards,
Somsubhro

Anyone? :slight_smile: :pray:

Hello somsubhro,
How do you consider the highlighted cracks differently than the others, if I may ask?
Bob

Hello Bob @smith_robertj ,
Thank you for your response.

I consider any feature that is vertical and thin (basically resembling a crack) a crack. There appear to be a lot of cracks in the image, I have highlighted only a few. I understand that I may not be able to capture all of the visible cracks, therefore highlighted some of the more prominent ones.

Thank you.
Kind regards,
Somsubhro

Hello again somsubhro,
I would start by inverting the image making the cracks bright on dark, then use the Analyze > Directionality tool and Ridge detection tool again.
I will also look into it a little deeper to see what else may help.
Bob

1 Like

Hello Bob @smith_robertj,

Thank you for the ideas. I’ll try this out now.

Thank you so much.
Kind regards,
Somsubhro

Being the curious creature I am, were you able to determine anything to your satisfaction?

Bob

Hello Bob,

Indeed it did. Your ideas helped me head into the right direction, and then I used a couple of other plugins to finally arrive at a cleaner image (Anisostropic diffusion, directionality-based morpholibj plugins, and shape filter).

Thank you.
Som

1 Like

There is no doubt that your determination was the key factor.
Congradulations
Bob

1 Like