This is in fact one of the suggested ways to implement non-linear transformations on objects. In “An Introduction To Ray Tracing” there’s a short blurb on implementing operations such as a twist or bend as the inverse of such a thing on the ray.

The only real problem with this is that the intersection function becomes a lot more complicated to compute and so takes a lot longer. But if you’re willing to wait, this is a great solution (which isn’t too hard to implement) to getting mathematically perfect non-linear transformations.

So yes, by applying these transformations to shadow rays, you could end up with some fun-shaped shadows. Might even be fun to have a bit of randomness added with some soft shadows.

Abstract
In this paper, general relativistic ray tracing is presented as a tool for
gravitational physics. It is shown how standard three-dimensional
ray tracing can be extended to allow for general relativistic visualization.
This visualization technique provides images as seen by
an observer under the influence of a gravitational field and allows
to probe spacetime by null geodesics. Moreover, a technique is
proposed for visualizing the caustic surfaces generated by a gravitational
lens. The suitability of general relativistic ray tracing is
demonstrated by means of two examples, namely the visualization
of the rigidly rotating disk of dust and the warp drive metric.

One aspect where this may be interesting is in physic: using appropriate curves you could build a raytracer that takes relativistic effects into consideration: when a ray approaches a massive object (i.e a star), the space distortion modifies the trajectory of the rays, wich becomes curve (in the 3d space). I think that someone did something similar a few years ago, by simulating how an observer would see the universe while falling in a black hole. They rendered a short movie, but I don’t remember who did that.
cignox1 is offline Report Bad Post Mark this post as useful
and
yarniso: (http://forums.cgsociety.org/member.php?u=90993)
—————–
There have indeed been previous attempts to include relativity into ray tracing, providing some interesting results. Check for example:

http://www.hakenberg.de/diffgeo/special_relativity.htm
http://www.anu.edu.au/Physics/Searle/
http://www.fourmilab.ch/cship/cship.html

Enjoy
cignox1: (http://forums.cgsociety.org/member.php?u=209425)
======================
here a few other ones:

http://www.photon.at/~werner/Relativity.html
http://www.photon.at/~werner/black-earth/
http://casa.colorado.edu/~ajsh/schw.shtml
==============
so who ever is interested you can have a look at provided links.