I recall from a section at CS61a that taking care of the refractive rays are somewhat unintuitive and difficult. I can also see why this is heavy in computation; we must do a ray tracing for every batch in the image plane and track its path that we do not know how long.
When animators model specific vintage lens effects, would they have to model all the internal elements of the lens for ray tracing, or is there a globalized way of rendering the lens affected image? I could also imagine that the combined effect of the lens could be modeled as a sequence of transforms that could be combined into a single transform that wouldn't meaningfully affect the runtime of the algorithm. There's a very cool video about some of the attention to detail in animation at pixar in modeling specific vintage lenses and filters! https://www.youtube.com/watch?v=AcZ2OY5-TeM
What happens if there are two mirrors next to each other? Do rays bounce forever or is there some ray "absorbtion" in surfaces?
Thanks for sharing that video. The split focus diopter is a mechanism I had never heard (beginning of video). It's particularly interesting how lifelike camerawork enters animation films. These concepts are easier to learn about here because they are paired with recognizable films. Great share.
@gavinmak, I'm pretty sure that it's impossible for a ray of light that bounces between two mirrors to ever get through the aperture of a camera. Practically speaking, in ray tracing I think there's a limit to how many bounces you compute out for.