A bit of a digression, but since it came up... I did quite a bit of this kind of thing at StereoD, where understandably relationships between positions and cameras is paramount.
The VFX team there has a pretty good toolset forming of custom ICE nodes to deal with just this kind of thing, which started out with simple compounds similar to the one Vincent has shared. For instance, if you project a particle for each pixel of a stereo depth map down the frustrum of a camera you have the beginnings of a 3d visualization of the depth of a scene (though it's complicated by the fact that StereoD has learned to use depth as a creative medium and therefore a strict linear interpretation of a depth map doesn't reflect what their stereo artists do to enhance a shot any more than color in a shot is a 1:1 interpretation after a shot is graded.) Stereo and ICE present some cool challenges by the way: with stereo information like a depth map brought into ICE it becomes possible to have your ICE sims interact,collide, and be occluded with footage. We often talk about the down sides of stereo but that extra depth information has some huge implications for VFX and StereoD was a great place to explore some of that. Going back to 2d work I find I particularly miss not having a depth map available to occlude particles as they go "behind" footage elements.
