Multilayer Light Simulations: More Beautiful Images, Faster


Dear Fellow Scholars, this is Two Minute Papers
with Károly Zsolnai-Fehér. Today we are going to talk about the craft
of simulating rays of light to create beautiful images, just like the ones you see here. And when I say simulating rays of light, I
mean not a few, but millions and millions of light rays need to be computed, alongside
with how they get absorbed or scattered off of our objects in a virtual scene. Initially, we start out with a really noisy
image, and as we add more rays, the image gets clearer and clearer over time. The time it takes for these images to clean
up depends on the complexity of the geometry and our material models, and one thing is
for sure: rendering materials that have multiple layers is a nightmare. This paper introduces an amazing new multilayer
material model to address that. Here you see an example where we are able
to stack together transparent and translucent layers to synthesize a really lifelike scratched
metal material with water droplets. Also, have a look at these gorgeous materials. And note that these are all virtual materials
that are simulated using physics and computer graphics. Isn’t this incredible? However, some of you Fellow Scholars remember
that we talked about multilayered materials before. So what’s new here? This new method supports more advanced material
models that previous techniques were either unable to simulate, or took too long to do
so. But that’s not all – have a look here. What you see is an equal-time comparison,
which means that if we run the new technique against the older methods for the same amount
of time, it is easy to see that we will have much less noise in our output image. This means that the images clear up quicker
and we can produce them in less time. It also supports my favorite, multiple importance
sampling, an aggressive noise-reduction technique by Eric Veach which is arguably one of the
greatest inventions ever in light transport research. This ensures that for more difficult scenes,
the images clean up much, much faster and has a beautiful and simple mathematical formulation. Super happy to see that it also earned him
a technical Oscar award a few years ago. If you are enjoying learning about light transport,
make sure to check out my course on this topic at the Technical University of Vienna. I still teach this at the University for 20
Master students at a time and thought that the teachings shouldn’t only be available
for a lucky few people who can afford a college education. Clearly, the teachings should be available
for everyone, so we recorded it and put it online, and now everyone can watch it, free
of charge. I was quite stunned to see that more than
10 thousand people decided to start it, so make sure to give it a go if you’re interested! And just one more thing: as you are listening
to this episode, I am holding a talk at the EU’s Political Strategy Centre. And the objective of this talk is to inform
political decisionmakers about the state of the art in AI so they can make more informed
decisions for us. Thanks for watching and for your generous
support, and I’ll see you next time!

Dereck Turner

Leave a Reply

Your email address will not be published. Required fields are marked *