Segmenting low SNR fluorescent image

Hi all,
I need to segment bone dyed with alizarin red from a dark background.
9dpa - wt9 - hq.tif (14.4 MB)
I have been playing around a fair bit with “non local means denoise” and “enhance contrast” in imagej. The results weren’t bad but thresholding or edge detection didn’t get me where I wanted. That is a problem because I sort of pitched my PI with how this is all going to be “semi automated”.
Right now I’m glancing at marker controlled watershed but I suppose I ought to look for help first. Any input would be greatly appreciated !