Visualizing three-dimensional root networks using computed tomography

We describe a method to reconstruct the root network from a three-dimensional image generated with computed tomography. The X-ray absorption of roots and the interface of the air–soil material is very similar and the contrast extremely low. The method consists of three steps: contrast enhancement using a non-linear diffusion filter, thresholding based on Rosin's method and extraction of the main features using a morphological connectivity algorithm. With this approach, we are also capable of reconstructing fine roots with high fidelity, which is shown with a performance experiment using artificial images with increasing noise. The method is applied to an X-ray microtomography data set of two alder roots, grown for 4 months in a natural moraine soil. To simplify the soil material, the bare roots were transferred into a Plexiglas receptacle filled with quartz sand. The reconstructed image is compared with a photograph of the real roots.