Chapter 3. Related Work
Despite the advanced tools for 3-dimensional tree generation available today, there is still research and further development being done. In this chapter, some recent research results will be reviewed. The reviewed papers are relevant for this thesis as a background. Some discussed papers deal with the generation of a tree or parts of a tree, while others are important from a technological point of view.
3.1 Bark Synthesis.
An important aspect of realistic-looking trees is realistic bark. Current tree generators usually create bark by applying texture maps. These texture maps can be used as bump mapping or displacement mapping to increase the visual quality of a 3-dimensional tree model. The problem of using texture maps arises when two branches have to be connected. This can create visual artifacts at the connection points created by texture differences on each branch. Usually, each branch is assigned a texture map, which is applied using a cylindrical projection. The top of the texture of one branch meets the bottom part of a different texture on another branch. The conjunction of these two textures results in a seam. In real trees, such seams are not present, since bark texture grows seamlessly and organically.
“Automatic tools use tillable textures to represent bark, which give unrealistic results on trees (i.e. No continuity at branching, no control of the features, and stretching depending on the branch size).”(Lefebvre and Neyret 2002). This is reasonable for situations when a tree is rendered from some distance, but it is not sufficient for very close views of a tree.
Another reason is that a tree trunk is usually wider at its base and close to its roots up higher it becomes narrower. The same features can be observed in branches. Applying texture on such an irregular cylinder makes the bark at the top part of the cylinder look compressed. The most straightforward way of dealing with such artifacts is by drawing bark textures manually. This can be a very labor intensive task requiring a great deal of manual work that could be automated.
Sylvain Lefebvre and Fabrice Neyret, in a joint research project, addressed this problem and came up with inspiring results. Their approach generates either a texture or mesh based upon the bark. This proposed method overcomes problems of seams in bark and trunk conjunctions. This method also doesn’t produce a compressed bark look in the places where the diameter of the trunk model is wider. Lefebvre and Neyret’s approach creates a realistic look for fracture-based barks.
Bark is a result of the tree’s growth process, which grows a new layer each year or each rainy season. New layers of bark are placed inwards, meaning the outside of the bark is the oldest layer, and the newest layers are closest to the wood tissue. The older the tree, the more layers grow on top of old ones. As the circumference of outward layers increases, fractures appear. The outer layers crack as a result of the diameter growing and its increased pressure on outer bark layers.
In their paper, the authors suggest to divide the trunk into transverse strips, which could be described in a 1-dimensional space for simplicity. Then the growth of the trunk or branch would be transferred to increase the length of these strips. The cracks are generated in these strips and then the original strips are covered with bark texture and the cracked ones are filled with crack texture. It could be either texture or 3-dimensional models that are inserted. The position of the crack’s appearance is aligned to the same crack in the next strip, this way the cracks are aligned along the trunk’s direction. Since the trunk or the branch has a different radius at different heights, the proposed model must take these differences into consideration. Otherwise the result would be a compressed texture appearance where the trunk or branch width is smaller, and stretched where it is wider. Therefore strips must represent the real circumference of the branch or trunk. In wide trunk areas we have long strips and in narrow areas short ones. Then the generation of cracks takes place.
Fig 21. Picture taken from: Sylvain Lefebvre Fabrice Neyret. 2002. Synthesizing Bark. ACM International Conference Proceeding Series; Vol. 28
As Figure 21 Illustrates, the bark breaks in its weakest segment in this model. The authors create a weakness map in their approach, and combine it with random position generations. This bark synthesis model achieves convincing results. The bark looks quite realistic compared to previous automating approaches that use simple texture tiling. Figure 22 shows the results of an implemented bark synthesis model.
Fig 22.Picture taken from: Sylvain Lefebvre Fabrice Neyret. 2002. Synthesizing Bark. ACM International Conference Proceeding Series; Vol. 28
The paper discussed above is very important in tree modeling, since it deals with bark as a 3-dimensional object and tries to simulate realistic-looking bark. This is a promising approach in bark simulation and generation.
you can also wisit my portfolio.
dont like english? u can (try) readig it in lithuanian or anny other language here! alternative, has no lithuanian :(