grain size measurements

[Peter Oslanec-2]

Hello.

I am looking for some advice for grain diameter measurements (ferets
diameter). Example of micrograph is
here<http://www.materialing.com/files/upload/example.jpg>.
Are there any ways how to determine the grain size? I mean som technique
based on tresholding, profiles, some math operation on source picture, not
clicking each grain on micrograph. I don't want exact solution, only some
advices, maybe some ideas. Thanks

Best regards Peto.

--
Peter Oslanec
Institute of materials and machine mechanics
Slovak Academy of Sciences
Racianska  75
831 02 Bratislava
Slovakia

[Michael Schmid]

Hi Peter,

this looks almost like an impossible task. Even with the eye and my  
highly parallel neuronal net behind it, trained over several  
decades ;-) in some places I can hardly tell what is a bona fide  
grain boundary and what isn't.

I guess that you have to
(1) get rid of the polishing scratches (not only the deep horizontal  
ones but also those under an angle, looking pretty much like a grain  
boundary, only straighter)
(2) use more magnification and avoid out-of-focus areas like on the top,
(3) use stronger etching and/or better (more grazing?) illumination  
to get more contrast.

Then, you could try by some kind of Laplace filter to enhance the  
lines (convert to 32-bit, convolve with a matrix having a high  
positive value in the center and negative values summing up to the  
center value around it).

For finding the edges of the grains having different brightness than  
the others, you could also try to first get rid of the etched lines  
(e.g. Process>Noise>Remove Outliers), then using the Thresholded Blur  
from the documentation wiki,
http://imagejdocu.tudor.lu/doku.php?
id=plugin:filter:thresholded_blur:start
to make the grains more uniform (but keep the sharp pedges) and then  
Process>Find Edges.
You can then combine these segmentation lines with the etched lines,  
e.g. by the Image Calculator Plus or Blend Images plugin from the  
ImageJ Web site.

Finally, if there are no breaks in the lines, thresholding and  
"Analyze particles" with "fill holes" should work.

Michael
________________________________________________________________

On 19 Nov 2009, at 15:55, Peter Oslanec wrote:

> Hello.
>
> I am looking for some advice for grain diameter measurements (ferets
> diameter). Example of micrograph is
> here<http://www.materialing.com/files/upload/example.jpg>.
> Are there any ways how to determine the grain size? I mean som  
> technique
> based on tresholding, profiles, some math operation on source  
> picture, not
> clicking each grain on micrograph. I don't want exact solution,  
> only some
> advices, maybe some ideas. Thanks
>
> Best regards Peto.
>
> --
> Peter Oslanec
> Institute of materials and machine mechanics
> Slovak Academy of Sciences
> Racianska  75
> 831 02 Bratislava
> Slovakia

[Peter Oslanec-2]

Hi Michael,

thanks for your advices. They are very useful, but I have got some
limitations.

(1) Polishing scratches: I can deal with finer polishing maybe on (let's
say) several micrographs but I need to process tens of micrographs. This is
quite time consuming. But yes, this can be done. The deep horizontal ones
are not scratches, these are also boundaries, but some other type like
inherent grain boundaries.

(2) Technically I am using biggest magnification on our optical microscope
and yes, the out of focus areas can be avoided. But it is problem of sample
preparation. When dealing with quite soft material, the corners are always
on out of focus plane at big magnifications (1000x) (at my conditions of
sample preparation).

(3) With stronger etching I will get overetched microstructure which is
quite "unreadable" also for me.

But thanks for your mail, I was able to confront my ideas with somebody who
is working at this field. Maybe I won't be able to this with automated
processing.

With best regards

Peter.
2009/11/20 Michael Schmid <[hidden email]>

> Hi Peter,
>
> this looks almost like an impossible task. Even with the eye and my highly
> parallel neuronal net behind it, trained over several decades ;-) in some
> places I can hardly tell what is a bona fide grain boundary and what isn't.
>
> I guess that you have to
> (1) get rid of the polishing scratches (not only the deep horizontal ones
> but also those under an angle, looking pretty much like a grain boundary,
> only straighter)
> (2) use more magnification and avoid out-of-focus areas like on the top,
> (3) use stronger etching and/or better (more grazing?) illumination to get
> more contrast.
>
> Then, you could try by some kind of Laplace filter to enhance the lines
> (convert to 32-bit, convolve with a matrix having a high positive value in
> the center and negative values summing up to the center value around it).
>
> For finding the edges of the grains having different brightness than the
> others, you could also try to first get rid of the etched lines (e.g.
> Process>Noise>Remove Outliers), then using the Thresholded Blur from the
> documentation wiki,
> http://imagejdocu.tudor.lu/doku.php
> ?id=plugin:filter:thresholded_blur:start
> to make the grains more uniform (but keep the sharp pedges) and then
> Process>Find Edges.
> You can then combine these segmentation lines with the etched lines, e.g.
> by the Image Calculator Plus or Blend Images plugin from the ImageJ Web
> site.
>
> Finally, if there are no breaks in the lines, thresholding and "Analyze
> particles" with "fill holes" should work.
>
> Michael
> ________________________________________________________________
>
>
> On 19 Nov 2009, at 15:55, Peter Oslanec wrote:
>
>  Hello.
>>
>> I am looking for some advice for grain diameter measurements (ferets
>> diameter). Example of micrograph is
>> here<http://www.materialing.com/files/upload/example.jpg>.
>>
>> Are there any ways how to determine the grain size? I mean som technique
>> based on tresholding, profiles, some math operation on source picture, not
>> clicking each grain on micrograph. I don't want exact solution, only some
>> advices, maybe some ideas. Thanks
>>
>> Best regards Peto.
>>
>> --
>> Peter Oslanec
>> Institute of materials and machine mechanics
>> Slovak Academy of Sciences
>> Racianska  75
>> 831 02 Bratislava
>> Slovakia
>>
>

--
Peter Oslanec
Institute of materials and machine mechanics
Slovak Academy of Sciences
Racianska  75
831 02 Bratislava
Slovakia