Help with segmentation of EM images

Does someone have ideas how to segment the round structures in this image?
I was trying for some time with the “Trainable Segmentation” but not with too much success. I feel one somehow has to add the information that one is looking for round objects, but I don’t know how to do this. Any ideas?


the image has an interesting histogram. How comes?

With respect to your question, you may have a look at the “Hough-Transformation” for circular objects. There is at least one ImageJ-plugin that does this transformation.



1 Like

Hi Herbie,
What do you mean by funny histogram? It is an EM image not fluorescence…
I tried this:
But could not get it to work; I feel there might be too many other edges in my image that are not part of the object boundaries.

Well Christian,

I didn’t write funny histogram but interesting. It has regularly missing values and it fits a Gaussian very well. Both is remarkable.

The spatial resolution of the image is poor. Do you have higher resolved versions?

If so, one could try to smooth the image and then find edges …



Hi @Christian_Tischer

That’s true. I would suggest Difference of Gaussian. DoG finds round structures because the filter is round.

Do you have this image in high resolution?

I changed the uploaded image to a higher resolution.

Is this the original file? A good image analysis is very image quality dependent. That’s why you will get the best results with the raw image data.

I guess not, it was sent to me by a collaborator; do you think the quality is not good enough?

I think the resolution is really low. If you take a look at the border of the circles you can see that it is not even a pixel wide at all positions. I hoped that this will change with a higher resolution.