Isolating and measuring faint droplet paths in images

Hello. OpenCV newbie here. We have a task to extract and measure droplet traces shown here as the faint lines that are barely out of the background noise. (See near vertical droplet in centre of attached image). I’ve been trying various techniques but its proving extremely difficult. The image has already been through openCV contrast increasing process as the droplets are not visible at all in the original images. Canny and edge extraction just produce too many results. Any suggestions on techniques to investigate would be greatly appreciated. Many thanks.

maybe a bit more lowpass

definitely try locally adaptive threshold. theshold level should be an absolute quantity over/under the mean/median, not relative to the local neighborhood’s stddev or anything.

do the black pieces of debris mean anything? can you get rid of them physically?