Hi guys. Does anyone one how to do a segmentation process that generates non-overlapping quadratic gray tiles within the image? I’m running the ActinAnalyzer2D from Mitobo and it requieres such segmentation to run. For each tile four co-occurrence matrices are calculated, and then for each matrix 10 customary texture features are calculated. The features currently used are contrast, local homogeneity, entropy, correlation, auto- correlation, dissimilarity, cluster shade, cluster prominence and maximum probability. Unfortunately, the user’s guide in the mitobo plug-in doesn’t explain how to do this segmentation and further labeling, so that the ActinAnalyzer2D recognizes and compares each cell within the image. Here I show you both, the original image and the segmented-labeled image. Thank you everyone.
can’t read the images…try jpg or png.
The MorphoLibJ plugin has the morphological segmentation plugin shown above…
If you think multi feature statistical analysis will work (trainable machine learning)…try the Trainable Weka Segmentation plugin in FIJI. It lets you choose the features (like texture measures) yourself.