challenging patterns in SEM-image

[alexzive]

Hello everybody,

I am a new user of ImageJ and I have a problem with pattern identification in grayscale images.
I tried to solve the problem with Gimp (which I know better) but results are not satisfactory.
Referring to the attached image taken with a Scanning Electron Microscope
you can distinguish three regions:  big globular black cavities, thin black "cracks" and some "gray
thin cracks" (crack~line).

Now to my problem: how to segment the picture isolating only the "gray thin
cracks" ?

Problem: such gray thin cracks share some grayscale values with several
pixels in the remaining non-black part of the image and most of these pixels are
actually noise (this is a problem of SEM, I have also tryed to adjust
contrast/brightness during SEM scannning, but this does not solve the
problem).


Any idea/inputs would be very much appreciated!!

Many Thanks in advance,
Alex

[alexzive]

I forgot the picture, sorry

01c01_WL1.tif

[Albert Cardona]

alexzive wrote:

> Now to my problem: how to segment the picture isolating only the "gray thin
> cracks" ?
>
> Problem: such gray thin cracks share some grayscale values with several
> pixels in the remaining non-black part of the image and most of these pixels
> are
> actually noise (this is a problem of SEM, I have also tryed to adjust
> contrast/brightness during SEM scannning, but this does not solve the
> problem).
>  

Try the "connected thresholding":

http://imagejdocu.tudor.lu/doku.php?id=plugin:segmentation:thresholding_by_connection:start

If there are broad background illumination gradients, you should
consider running a high-pass filter first (try the FFT > Bandpass filter
with a 0, 50 window for example, where 50 is much larger than the
thickness of your thin cracks.)

Albert

--
Albert Cardona
http://albert.rierol.net