This image is a cross-section from a branch of a black spruce showing the xylem structure.
I am trying to distinguish latewood (the darker pink color at the edge of the tree ring) vs earlywood (the lighter color at the beginning of each tree ring). I also want to distinguish the inside of the cell (the white open space) from the cell wall, but have found that a separate classifier is needed for that.
This is my first time using TWS, but I have been playing around with it for about a week now. I read the ImageJ page on the settings for TWS (https://imagej.net/Trainable_Weka_Segmentation.html#Settings) and vaguely understood what it was talking about, but was hoping to get a better idea of what each filter did and if I should try to manipulate the membrane thickness, patch size, etc.
So my first questions are:
Does anyone know of a more detailed explanation for each of these settings that I could look at?
If not, then does anyone have any suggestions for what the best settings would be? For the most part, I’ve been having success, but images like the one above where there is very minimal earlywood do not have as clear results as I would like.
I have been opening my images with bio-formats, which opens the .tif files as three separate channels. I trained my classifier model to the green channel because it seems to work the best. Is this common practice, or is there an issue with this tactic?
And finally, when I save the classifier model, does anyone know what information is being saved? When I load the same classifier for a different picture, add more class samples, and re-train the classifier, what does that do?
Any information is appreciated! Thanks!