Image thresholding method that adapts to changing signal-to-noise levels?

I am looking for an image thresholding method that can be robust to differing signal-to-noise levels of an image. Under the same microscope imaging conditions, I may get certain fields of view that have higher background levels than others and some brighter signals than others. Can someone recommend an algorithm that is robust to these differences in image conditions? Thank you.

High SNR Image:

Low SNR image:

Answer 1

[Image Analyst]

First try a longer integration time or averaging a bunch of frames together to try to reduce noise (if the particles don't move). Then try some noise reduction,

https://users.soe.ucsc.edu/~milanfar/publications/journal/ModernTour.pdf

then try a tophat filter, imtophat().

They kind of look like astronomical images so you might try the Knox-Thompson method: https://www.google.com/#q=knox+thompson+noise or similar astronomical noise reduction methods.