New Scientist Article on Anisotropic Meshes

[nickej]

An interesting article on Science Daily on isotropic vs. anisotropic mesh generation claims that anisotropic meshes are more capable of accurately representing real-world objects. I assume that the article is referring to automatically generated meshes based on scans or data, but there’s a strong implication that it refers to hand-made meshes as well.

Considering that general 3D mesh construction technique emphasizes avoiding strongly anisotropic tris or quads as a given, is this a call to re-evaluate construction technique, or a case where something is theoretically better, but practically impossible or not useful?

Anyway, here’s the article: http://www.sciencedaily.com/releases/2013/11/131119112402.htm

[Zalamander]

Here’s the actual research. As it is often the case with research, it is more of a solution looking for a problem. The paper hints that it is better for certain types of simulation and that it leads to generally more efficient representations (no surprises there).
However, such meshes deform very badly compared to their “isotropic” counterparts. Both the “isotropic” and “anisotropic” mesh representations are completely unsuitable for direct user controlled modeling. The relevance to your modeling habits is none.