Alternative to threshold

[kirill8910]

Hello!

I'm trying to separate the cracks out (picture) as precise as possible. Grains on the background make ordinary threshold tricky. On other images that I have direction of cracks and grains could be the same, so Sobel like operators dont fit unfortunately.


Is it possible to provide labeling or segmenting according to cracks intensity change (cracks have at least sharp slope in the intensity change (plot profile))?

Could you suggest me any plugin or algorithm where I can set this kind of criteria for the segmentation.

Thank you!

[ctrueden]

Hi,

Is it possible to provide labeling or segmenting according to cracks
> intensity change (cracks have at least sharp slope in the intensity change
> (plot profile))?
>

Did you try "Find Edges" in the Process menu?

You could also try Fiji's Trainable Segmentation plugin for a teachable
algorithm.

-Curtis


On Wed, Dec 7, 2011 at 11:53 AM, kirill8910 <[hidden email]> wrote:

> Hello!
>
> I'm trying to separate the cracks out (picture) as precise as possible.
> Grains on the background make ordinary threshold tricky. On other images
> that I have direction of cracks and grains could be the same, so Sobel like
> operators dont fit unfortunately.
>
>
> Is it possible to provide labeling or segmenting according to cracks
> intensity change (cracks have at least sharp slope in the intensity change
> (plot profile))?
>
> Could you suggest me any plugin or algorithm where I can set this kind of
> criteria for the segmentation.
>
> Thank you!
>
> http://imagej.588099.n2.nabble.com/file/n7071422/cracks.png
>
> --
> View this message in context:
> http://imagej.588099.n2.nabble.com/Alternative-to-threshold-tp7071422p7071422.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>

[kirill8910]

Thank you for the response!

I tried to find edges but wooden annual rings on the background are distinguished as edges as well. I will try this trainable threshold then...

/Kirill

[Anderson, Charles (DNR)]

Kirill,

This will be very difficult because the cracks are not always darker or sharper or straighter than the annual rings. Your black arrowhead points to a crack within the first half of the transect, whereas the other end incorrectly links it to a pixel in the other half of the transect. The annual mark is the darker sharper one.

I think you will have to mark cracks manually, or resort to some specialized dendrochronology software to identify dark annuli, then remove them so that dark lines remaining are cracks.

If you google "ImageJ dendrochronology" you might be able to locate others using ImageJ for similar purposes, although I do not see any plugins specifically for this.  I face similar problems working with fish scales and otoliths.

Or, maybe make an acetate peel, where the cracks are ridges and everything else is smooth. Or maybe you could fill the cracks with white powder (flour?) and then identify them as the only white objects in images.

Charles Anderson

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Curtis Rueden
Sent: Wednesday, December 07, 2011 1:01 PM
To: [hidden email]
Subject: Re: Alternative to threshold

Hi,

Is it possible to provide labeling or segmenting according to cracks
> intensity change (cracks have at least sharp slope in the intensity change
> (plot profile))?
>

Did you try "Find Edges" in the Process menu?

You could also try Fiji's Trainable Segmentation plugin for a teachable
algorithm.

-Curtis


On Wed, Dec 7, 2011 at 11:53 AM, kirill8910 <[hidden email]> wrote:

> Hello!
>
> I'm trying to separate the cracks out (picture) as precise as possible.
> Grains on the background make ordinary threshold tricky. On other images
> that I have direction of cracks and grains could be the same, so Sobel like
> operators dont fit unfortunately.
>
>
> Is it possible to provide labeling or segmenting according to cracks
> intensity change (cracks have at least sharp slope in the intensity change
> (plot profile))?
>
> Could you suggest me any plugin or algorithm where I can set this kind of
> criteria for the segmentation.
>
> Thank you!
>
> http://imagej.588099.n2.nabble.com/file/n7071422/cracks.png
>
> --
> View this message in context:
> http://imagej.588099.n2.nabble.com/Alternative-to-threshold-tp7071422p7071422.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>

[Bob Loushin]

Have you looked into the Hough linear transform?  It can be tuned to the
crack width.  Look on the Plugins page of the ImageJ site and follow the
link, in the Links to External Sites section, to Wilhelm Burger and Mark
Burge's page.

Bob

-----Original Message-----
From: kirill8910
Sent: Wednesday, December 07, 2011 11:53 AM
To: [hidden email]
Subject: Alternative to threshold

Hello!

I'm trying to separate the cracks out (picture) as precise as possible.
Grains on the background make ordinary threshold tricky. On other images
that I have direction of cracks and grains could be the same, so Sobel like
operators dont fit unfortunately.


Is it possible to provide labeling or segmenting according to cracks
intensity change (cracks have at least sharp slope in the intensity change
(plot profile))?

Could you suggest me any plugin or algorithm where I can set this kind of
criteria for the segmentation.

Thank you!

http://imagej.588099.n2.nabble.com/file/n7071422/cracks.png

--
View this message in context:
http://imagej.588099.n2.nabble.com/Alternative-to-threshold-tp7071422p7071422.html
Sent from the ImageJ mailing list archive at Nabble.com.

[Olivier Burri]

kirill8910 wrote

I tried to find edges but wooden annual rings on the background are distinguished as edges as well.

Hello,

You should try the following procedure:
Try applying a median filter to remove your cracks (They are smaller than the annual rings so the filter will remove them). Then you can subtract this image from your original and it will make your cracks stand out.

Have a look at the results.


Best

Oli

[kirill8910]

Oliver, thank you for response!

I will try it, but I'm afraid the other thin annual rings will be removed also by median.

Kirill

[kirill8910]

Thanks for ideas Charles!

Imaging in dendrochronology sounds interesting, I will check it out!

Kirill

[kirill8910]

Thanks, Bob!

Unfortunately a huge part of cracks that I have are not straight lines. But worth trying anyway!

Kirill

[negupta]

Hello ! Recently I read one paper named 'Identifying microcracks in multi-phase crystalline rocks by X-ray CT' which claims to solve the problem of identifying cracks without using threshold method. For this purpose they wrote an algorithm used as Plugin in Image J software. I am trying to use that plugin but I am not able to find it. Does anybody have any idea about that plugin or where I can find it? Please let me know. Thanks, Neel