As we all know, when measuring the shapes (e.g. circularity) of objects with few pixels, strange things are happening. I think I saw a thread on this forum where these effects were discussed in some detail, but now I cannot find it anymore. Could one of you point me somewhere (could also be a publication) where these issues are discussed?
I read this recently:
I do not know if that answers part of your request.
Maybe you are referring to this post:
The key point in circularity is the measure of perimeter. Most methods results in a deviation from expected value, and the effect is increased for small objects.
As circularity depends on perimeter (on denominator), it is also impacted by discretisation.
For references, I can suggest the one I suggested in the thread (ITK implementation of perimeter and surface).
More recently, @iarganda also published a chapter dealing with these questions:
Arganda-Carreras, I. & Andrey, P.
Markaki, Y. & Harz, H. (Eds.)
Designing Image Analysis Pipelines in Light Microscopy: A Rational Approach
Light Microscopy: Methods and Protocols, Springer New York, 2017 , 185-207
Thank you both very very much!
Those were indeed the threads I was looking for!