Particle segmentation

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Particle segmentation

James Wesley-Smith
Me again.

To familiarise myself further with FIJI,  I deliberately chose gold
nanorods for their electron opacity and thus contrast, but I am struggling
to segment them. Thresholding successfully  separates them from the
background, but I have tried various settings of erode / dilate and open /
close routines to no avail - just cannot separate them completely. I even
looked for watershed segmentation, which tends to "over segment". I also
looked at answers to previous questions posted on this listserver without
much success.

 

Where am I going wrong? What segmentation steps would you suggest I follow
for this sort of sample (attached)?

 

 

Thanks again!

 

 

Best regards,

James

 


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

Test nanorods.tif (1M) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Particle segmentation

Michael Schmid-3
Hi James,

you can try the Adjustable Watershed. It does not get it 100% correct,
but most of the rods are separated correctly.

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

As a preprocessing step (before thresholding) you could duplicate the
image, run Find Edges, and add it to the original. This makes the edges
stand out more clearly. Then, after thresholding, the Adjustable
Watershed gets it almost perfect.

I also tried various methods based on grayscale (edge-preserving blur
plus Find Maxima, which can also do a segmentation). In several places,
the contrast within the particles is higher than the contrast between
the particle and the gap in between two particles, so this does not help.

For a perfect result, I fear that one would need some algorithm that
takes the expected shape of the nanorods into account. (I am not aware
of any; maybe others know more.)


Michael
________________________________________________________________
On 22.01.19 08:30, Jaime Wesley-Smith wrote:

> Me again.
>
> To familiarise myself further with FIJI,  I deliberately chose gold
> nanorods for their electron opacity and thus contrast, but I am struggling
> to segment them. Thresholding successfully  separates them from the
> background, but I have tried various settings of erode / dilate and open /
> close routines to no avail - just cannot separate them completely. I even
> looked for watershed segmentation, which tends to "over segment". I also
> looked at answers to previous questions posted on this listserver without
> much success.
>
> Where am I going wrong? What segmentation steps would you suggest I follow
> for this sort of sample (attached)?
>
> Thanks again!
>
> Best regards,
>
> James

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: Particle segmentation

Michael Schmid-3
Dear James,

In your image, there is an outlier pixel at position (667, 281), with a
pixel value of 171 (the surrounding pixels are in the range 70-80).

This pixel acts as a seed for a dividing line of the Adjustable Watershed.

You can use, e.g. Process>Noise>Remove Outliers. For single outlier
pixels, a radius of 0.5 is enough; and specify a sufficiently large
threshold to avoid affecting the gaps between the particles, e.g. 50.

[BTW, please make sure you reply to the correct thread; otherwise it is
impossible to find things in the list archives]

Best,

Michael
________________________________________________________________

On 22.01.19 12:52, Jaime Wesley-Smith wrote:
 > Dear Michael
 > I think the Adjustable Watershed does a pretty nice job (with the
exception
 > of particles 23-25) on the image attached! Hope I have recorded the macro
 > correctly for you to see...
 >
 > I will also try your suggestion of the overlay.
 >
 > Many thanks again!
 >
 > Best regards,
 >
 > James

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: Particle segmentation

James Wesley-Smith
Thanks Michael, will pay more attention to the pixel values in future.

I didn't want to burden the listserver, but you are right- archive!

Thanks again.

Best regards,

James



-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
Michael Schmid
Sent: Tuesday, January 22, 2019 7:05 PM
To: [hidden email]
Subject: Re: Particle segmentation

Dear James,

In your image, there is an outlier pixel at position (667, 281), with a
pixel value of 171 (the surrounding pixels are in the range 70-80).

This pixel acts as a seed for a dividing line of the Adjustable Watershed.

You can use, e.g. Process>Noise>Remove Outliers. For single outlier
pixels, a radius of 0.5 is enough; and specify a sufficiently large
threshold to avoid affecting the gaps between the particles, e.g. 50.

[BTW, please make sure you reply to the correct thread; otherwise it is
impossible to find things in the list archives]

Best,

Michael
________________________________________________________________

On 22.01.19 12:52, Jaime Wesley-Smith wrote:
 > Dear Michael
 > I think the Adjustable Watershed does a pretty nice job (with the
exception  > of particles 23-25) on the image attached! Hope I have
recorded the macro  > correctly for you to see...
 >
 > I will also try your suggestion of the overlay.
 >
 > Many thanks again!
 >
 > Best regards,
 >
 > James

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html