I'm especially interested in the segmentation of haematoxylin-stained nuclei. I have used grayscale opening by reconstruction to homogenize the intensities, followed by marker-controlled watershed segmentation.

But there are cases where the haematoxylin staining is so poor that a wider structuring element is required for the reconstruction. This, as expected, leads to under-segmentation since most nuclei are surrounded by other nuclei.

An example image is attached. Any pointers?

Similar questions and discussions