Lighting and Motion
Problem
1: Lighting
Description
Animations of signs were difficult to understand or to
imitate properly because Paula had no shadows and the
viewer was not getting any depth or distance information.
The lack of shadows decreased the understanding of where
hands were in space and how near or far from the body
they were. This affected the reading and learning of the
signs, especially for those new to ASL.
Solution
To solve the depth problem, we changed the lighting setup
in the scene and included shadows. We also changed the
color of Paula's sweater to make shadows easier to see.
We
positioned the lights in the scene as shown below:
–
Omni lights emit light in all directions, like a candle
flame
–
Directional lights emit a column of light in a specific
direction, much like a flashlight.
The
directional light with the green asterisk acts as our
main light and is the only light to cast shadows. The
other lights act as “fillers” so no part of Paula's face
or sweater are too dark.
View
video of before and
after lighting adjustments.
Problem
2 : Unnaturalness of motion
Description
Although Paula's movements were accurate, the motion itself
seemed unnatural because it was clean and crisp at every
frame. The perceptual effect of motion over time needed
to be captured more correctly. Therefore, we explored
motion blur. In animation, motion blur is able to add
the visual impact of motion by involving some form of
blurring.
Solution
For this problem, we investigated Object Motion Blur and
Image Motion Blur. In the rendering software we used,
object Motion Blur is dependent on geometry, and a number
of samples are taken at each frame, and are then averaged.
In Image Motion Blur, the pixels are blurred after each
frame is rendered and so is not dependent on geometry.
In
addition to creating a visually pleasing blur for Paula's
movement, keeping the rendering time to a minimum was
a concern. We produced very satisfactory parameters for
Object Motion Blur and Image Motion Blur. The difference
was that the rendering time with Image Motion Blur was
about half of Object Motion Blur. Thus, we started to
produce the animations using Image Motion Blur. However,
the system we were using could not handle the Image Motion
Blur setup for large amounts of renderings and so we resorted
to using Object Motion Blur during rendering. It took
about 34 hours to produce 537 animations.
Now,
Paula's motion is accentuated by a slight blur depending
on the speed of movement and is a definite improvement
in the naturalness of the animations.
Study
conducted by: Tahseen Basheeruddin and Irena Svidovsky
Research supervisors: Dr. John McDonald and Dr. Jacob
Furst
Research
supported by a grant from The Computing Research Association
Committee on the Status of Women in Computing Research
(CRAW) under a program entitled Collaborative Research
Experience for women (CREW). Funding for this grant was
provided by the National Science Foundation's partnership
for advanced computational infrastructure's education,
outreach and training program (EOT-PACI) and by USENIX.
