Synthesizing Bark

Sylvain Lefebvre, Fabrice Neyret


Synthesizing Bark     PDF, 4.8 MB PS, 5 MB
Sylvain Lefebvre, Fabrice Neyret
13th Eurographics Workshop on Rendering 2002


Applications like landscape management, special effects and video games need to represent trees from far to close viewpoints. There exists sophisticated softwares like AMAP or XFrog that are able to produce tree models but none of these softwares handle the creation of bark details. They generally propose a standard texture mapping with a real bark sample. This is not sufficient for close viewpoints as there is no variation of appearence between branches. The bark appearence is simply scaled and that does not correspond to what can be observed in nature. Moreover, there often is no continuity at branchings and the texture sample is distorded due to the radius variation along the trunk.

As the bark has no influence on the tree growth, the problem of representing bark is a dressing problem: we need to paint details on the tree surface. Among the wide variety of bark that can be observed in nature, fracture based bark is one of the most common. For this reason, our model is dedicated to the representation of fracture based bark.

Our bark model simulates the propagation of fractures along a tree trunk and the enlargement of fractures with time. Below you can see samples of 2D simulation data.

We run the simulation directly on the tree geometry by building material strips around the trunk. Each strip can be fractured. A fracture can propagate from one strip to another. The complexity of the appearence is obtained by introducing phenomenological laws to control the behavior of fractures. These laws comes from our observation of real barks.

After doing the simulation we dress the raw data with details by texturing the epidermis surface and tear interior.

Our model is able to handle branching by doing the propagation of fractures in world space. Instead of propagating the fracture to the neighbouring strip of the same trunk, we propagate it to the neighbouring strip on the branch.

Results with various parameters

Some results of the branching algoritm

The model can run on top of a biological simulation of tree growth (like AMAP). In this case, we know the growth information and the model simply runs on top of the growing geometry (see video 1). But our model can also be used to create a texture for an existing tree model (such for applications like special effects or video games). As we do not know the growing information and as we do not want to deform the final geometry, we adapt our model so that it does a reverse simulation. Instead of growing the geometry the bark is shrinked (see video 2).
The last video shows the continuity of our bark texture around the trunk. See how the epidermis length is conserved: there is the same circular length of epidermis on top of the trunk as on the bottom of the trunk.

Here are some bugs of fracture propagation that result in interesting patterns ;^)

Real barks pictures.

All pictures and videos of this page are under copyright.