Having a hard time persuading imageJ to pick out objects inside of bacterial colonies for particle analysis because the variable background/object colour. Thresholding tends to grab a lot of unwanted bits while missing some objects. Any ideas how to improve?test1.tif (1.7 MB)
I’m not exactly sure what methods/tools you have tried thus far… but I would steer you towards trying Trainable Weka Segmentation. What are you trying to segment in this image in any case - the oval, dark areas?
Ideally I want to be able to separate out the dark peripheral objects (crystals) as well as the lighter central objects from the background of the colony so as to compute some kind of overall area or ratio. I’m failing with the particle analysis because I can’t neatly threshold everything I want into a binary image because of the variability in colours. I’m not very savvy with ImageJ but I’m basically looking for a more elegant method than drawing freehand circles around everything and measuring the areas.
If what https://forum.image.sc/t/help-with-particle-analysis-for-objects-inside-bacterial-colonies/27525/2?u=smith_robertj etarena, doesn’t work well then I would suggest k-means clustering.
You could also try edge detection to trace the outlines of the objects. Getting the binary blobs would then just be a matter of filling in the edge lines. There are some more sophisticated edge detection algorithms discussed on this forum:
Trainable Weka Segmentation works pretty well, thanks for the suggestion!