The problem is that the black cells starts to grow into each other and the walls disappears (red lines in picture). Is there a way to inflate the black cells without growing them into each other so that the cell topology is preserved? I just want thinner walls. The result doesn’t need to look exactly as the initial image.
I have the problem of finding a good thresholding value on the original CT volume: My thresholding either removes visual walls between the cells or fills up small cells. Could there be a way to postprocess my initial image (for example sharpening edges/shapes?) so I can make a binary image that looks like how we humans see the cells and walls?
Possibly look into using watershed based segmentation. Take a look at this jupyter notebook, segmentation of bacteria from a 3D Focused Ion Beam Scanning Electron Microscopy. Specifically the section titled “Seed based watershed segmentation”.
Maybe the binary thinning filter might be of interest to you - it keeps all cell walls, but you will get al of them thinned down to a single pixel.
In my experience it does not work too well if the structures to be thinned are too thick in the input image; and we have encountered what seems to be a bug in the current version, leading to additional artefacts in the thinning - but I don’t have the details of that issue handy at the moment.
