Distinction of background and object of interest

[skytalker]

Hi , I have a problem of separation the objects of interest from the background. Attach file is the sample image. I have tried to use some methods such as " thresholding " or followed the particle analysis example in the Vision example. However, these mothods were not work in this application, because the resulting of above methods distroted the shape of the objects. In this application, the objects of interest are the particle in line shape, as they are marked numbering in the attach files, these particle can not be distroted, because after the separation they will be measured. If you have any idea or tools please guide me.

Please advise.

Skytalker.

[Robert_Eastlund]

Please upload an uncompressed sample image.

[skytalker]

Hi ,

This is the uncompressed images.

regards,

[BCho]

Hey skytalker,

I can see the difficulty you are having separating the particles of interest from the background. This likely due to the fact that your region of interest is so close in pixel value to other objects in your image. The irregular shape and fact that the particles you care about are so different from each other. Out of curiosity, what sort of things are you looking to measure if you were able to completely separate the particles from the background? If it was known what you were interested in measuring in the end, it would help figure out how to go about filtering your particles.

-Ben 

[skytalker]

Hi Ben,
My target of this application is to measure the length and the thickness of (line shape) particles.  After the particles can be separated from its background i plan to use tool boxes such as Edge detection and Caliper to measure both thickness and the length of the particles.

Regards,

[Robert_Eastlund]

Skytalker,

Here is a Vision Assistant v8.5 script that aims at isolating the elongated objects in the image.  The script seems to work ok for the two sample images provided.  The attached .zip file includes the script and the two resulting images.  Let me know if the example script is isolating the correct particles.  It probably needs more work.

-Robert

[Vladimir_Drzik]

Robert,

Nice work. I wouldn't have said you could get such a good result with just Niblack thresholding and particle filtering.

Vladimir

[skytalker]

Hi Ben,
Yes, the resulting images in the .zip are the one that I expected to have. I also tried to run the script through the last step, but i cannot get the resulting as yours. Is there any more steps needed ?
Regards,

[Robert_Eastlund]

What version of Vision Assistant are you working with?

[skytalker]

Hi Ben,
I'm using Vision Assistance 8.6

Regards,