In conclusion, a semiautomatic 3D segmentation technique was proposed,
grounded on the immersion-based watershed algorithm. It was shown that
preprocessing the data with a 3D adaptive anisotropic diffusion filter has
a positive impact on the segmentation results. A posteriori merging of
basic volume primitives additionally reduces the user interaction time.
The proposed segmentation technique is successfully used to extract
quantitative volume information from 3D images of in vitro
as well as in vivo mouse and bird cerebella.
It has proved to be superior in comparison with existing invasive methods.
