Help adjusting picture for analysis of egg size

Dear fellow ImageJ’ers,

For my master thesis I am rearing black soldier flies on different diets, I am investigating how the diet influences reproduction. Therefore I am measuring the size of the egg clusters that the flies have layed. For this measurement I take a picture of an egg cluster and use freehand selection to get the size.
Now I can keep on doing it by hand, which is oke. But having 75 treatments, with each approx 40 to 50 egg clusters means I must analyse 3000 to 4000 pictures by hand. So if I can automate the measuring, it would save me a lot of time :stuck_out_tongue:
Example of the pictures I take:

The problem I face now, is that I cant seem to get a good contrast between the eggs and the rest of the image. Does anyone have an idea how I can adjust my image using ImageJ or the microscope which has settings to chance the RGB saturation?

Any bright ideas about with which parameters to fiddle?

Thanks in advance,

Seppe

Do you need to meaure what I assume are individual eggs or the cluster as a whole ?

@susan-sheehan is correct… if you are interested in the entire cluster, I would recommend a larger field-of-view if possible.

Too - for segmenting, as the eggs seem to be somewhat uniform in color, you can try using the Color Threshold feature. Perhaps this is a good place to start.

Hope this helps a bit. Feel free to post again with any follow-up questions, etc.

eta :slight_smile:

They are indeed individual eggs, but I need to measure the area of the entire cluster. As they are stacked on top of each other, counting the eggs from pictures is pretty much impossible. Weighing them is also pretty difficult and estimating the area of the cluster turns out to be pretty reliable.

I tried to balance the colour tresholds and contrast became alot higher. But im not sure how to continue from this point on. I counted the number and area of vessels in tomato plant using image j. This i could do fairly easy with the measure particles function with size x and shape y when I created a binary image. But I cant figure out something similar to this method for the egg cluster measuring.

Thx for the quick replies!

Seppe

Ill try some pictures tomorrow from further away.

As long as you can keep the lighting conditions consistent and similar across your dataset, you can try to train a SIOX Classifier (Plugins > Segmentation > SIOX: Simple Interactive Object Extraction) and apply it to all your images (Plugins > Segmentation > Apply saved SIOX segmentator).

I was getting a reasonable segmentation when quickly trying with your posted image:

Hope that helps.

3 Likes

Awesome! Thats the kind of trick I was looking for. Did not have any time to test it yet but im sure with a little help from google I should get it working.

thanks