Detecting low contrast spheres

Hi, I would like to analyse spheres from a SEM image where some spheres have quite low contrast. I have been experimenting with adjusting C&B, despeckle, enhance contrast and thresholding but the biggest issue is that when I threshold a part of the low contrast spheres are lost, therefore they do not get detected as spheres but only part of the rim. For the same reason “fill holes” does not help either. Ultimately I only want to detect the area/diameter of the spheres. Any advice?

Thanks in advance.

You can use a Laplacian filter like e.g. FeatureJ Laplacian to enhance large spots in your image, and then use the threshold:

Alternatively, use one of the spot detectors built into the TrackMate plugin.

1 Like

You can do a little pre processing step, search the mean of the entire image and rest it to the image, then drop negative values and repeat all your steps. You will have noting in the image but you spheres :wink:

Thank you for your answers. Bishopwolf, what do you mean by “rest it to the image”, subtract the the MGV from the whole image?

Yes, substract the mean value from the entire image and create a new one with only positive values, put the negatives to 0

try this, and optimize parameters:

run("Auto Local Threshold", "method=Niblack radius=300 parameter_1=2 parameter_2=0 white");
2 Likes