Graphics Reference
In-Depth Information
a traditional clay sculpting studio, we have artists who have made the move to 3D
and can easily handle any digital cosmetic surgery needed when scanning organic
subjects.
A decade ago, it wasn't commonplace to scan actors, but the production compa-
nies that make the movies have now realized how important this is to their vendors.
They might have a scene where the film production company has not yet hired a
visual effects vendor, even though filming is already occurring. The visual effects
supervisors know that the vendor is likely to need 3D reference several months later,
so they'll ask us to go ahead and scan anything and everything on set. Maybe they'll
need it andmaybe they won't, but it's neither feasible to rebuild a set nor get an actor
who's already filming another movie to come back and get in makeup and costume
for a texture shoot or 3D scan.
Once acquired, the scan data is often used by toy companies for action figure
design and by video game companies for movie tie-in games. Actors often voice
concernabout thedatabeingused to later animate themandmake them“sell vacuum
cleaners” after they die, but they probably aren't aware of the inevitability of purely
image-based 3D reconstruction algorithms that will make the whole issue moot. It's
conceivable that at some point in the near future, 3D information can be extracted
from even the earliest motion picture footage.
RJR: Do you also use patterns of structured light?
Chapman: Structured light systems can also work well for head and body scanning,
but they're riskier to use in practice compared to laser-stripe systems due to the
complexity of the underlying image processing. Often, an actor is only available for
a few short minutes and may even called back to the set in the middle of a 3D scan.
As one A-list actor put it after a few seconds of delay in starting the body scanning
on the set of a blockbuster, “Let's go, gentlemen!” As a result, many visual effects
supervisors will lean toward using laser-stripe systems for actor scanning simply
because they know they're much less prone to failure compared to structured light
devices.
Trying to project structured light onto larger objects is extremely difficult since you
can't get a projector bright enough and you can't get the camera far enough away to
make the algorithms work. There are also limitations you run into with the vibration
of the mounting systems and the resolution of the cameras. We have a structured
light system that could theoretically scan a jumbo jet, but realistically that's a job
much more suited to LiDAR.
We still use structured light for scanning our smallest subjects because it excels
at capturing fine details that laser “blooming” obliterates. For example, we used
structured light to capture a small pirate coin at sufficient detail for it to be projected
three stories high in the opening credits of a film. Real-time performance capture
scanning once required structured light solutions, but stereo matching has evolved
to replace that need, with the added benefit of not blinding the actor with a projector
pattern.
One common issue with structured light is that if you're scanning a human subject
and they are moving, as inevitably happens, you see a sort of a “cheese grater” effect,
where some parts of the surface will be further out or in than other parts, and you see
