Vector fields on meshes are useful for numerous things and hence they have many applications in computer graphics, such as texture synthesis, anisotropic shading, non-photorealistic rendering, hair grooming, surface parameterization, remeshing and many more. Sometimes they are quite easy to compute and sometimes it gets tricky. Depending on the task, we maybe have to work with an existing vector field but most of the time we need to generate a new one, typically based on a small set of user-defined constraints. Either way, we usually want the vector field to be tangent and more importantly, smoothly varying over the surface. Well, the tangent part is easy. One way how we could generate such a field is simply by computing the gradient of a scalar function, and the result is of course a tangent vector field. This is easy to do but the resulting field might not be what we want. If, for instance, we need to compute a vector field which should rotate around a point, it won’t be possible using the above mentioned method. The reason is quite simple – since it is a gradient field, it is per definition always rotation-free. In other words, curl is always zero. Of course, if we compute the orthogonal complement of each vector, in other words rotate it around 90 degree, we would get rotations. But at the same time we would naturally end up with a divergence-free vector field, which again might not be what we want.

Instead of utilizing the gradient, a much more flexible way to design vector fields is by prescribing user-defined constraints, which are then interpolated over the surface. These constraints are typically sinks, sources and vortices but could also be some edges on the mesh, which guide the direction of the field. Well, at this points things become a bit tricky but fortunately it is a very well researched topic and Google unveils the great wisdom hidden in many research papers. A paper I’ve particularly enjoyed is “Design of Tangent Vector Fields” by Fisher, Schröder, Desbrun and Hoppe. The described method is based on quadratic energy minimization and relies only on standard operators from discrete exterior calculus.

M. Fisher, P. Schröder, M. Desbrun, H. Hoppe: Design of Tangent Vector Fields

After reading the paper I did some prototyping in python and while it worked generally quite well it was of course rather slow. It was probably around 2014 at the time and I wanted to port the algorithm to the HDK or at least use a sparse matrix library in python but something intervened and later I lost somewhat interest in it. Well, now it’s 2017 and this blog seemed to be a good opportunity to dig out the old file. It´s still slow (even on a fast computer) but if you’re not in hurry it does a pretty good job.



Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s