CMU teams with Disney to refine human-like animation
-
A 3-D face model enables animators to create expressions such as a wink and a smirk. -
A region-based 3-D face model developed by Disney Research, Pittsburgh, and Carnegie Mellon University enables animators to add constraints, indicated by black markers on the model, to create expressions such as a wink and a smirk. -
A system developed by Disney Research Pittsburgh and Carnegie Mellon University uses body-mounted cameras to capture motion for use in animation. Here, cameras attached to a runner outdoors record motions used to render an animated runner. -
A system developed by Disney Research Pittsburgh and Carnegie Mellon University uses body-mounted cameras to capture motion for use in animation. Cameras attached to a runner outdoors record motions used to render an animated runner, shown here.
Share with others:
The appearance of reality is paramount in the world of make-believe, where the viewing public yearns for animated characters who walk, run, twist and turn like real people and display a full range of human emotions and expressions.
With those goals in mind, two research teams from Carnegie Mellon University and Disney Research Pittsburgh are busy developing computer techniques that quickly produce realistic animated facial expressions and body motions -- a real challenge given the myriad capabilities of the face's 43 muscles, the complexity of the human body and our ability to perceive subtle changes in human motion.
The facial expression and body movement projects are in the research-and-development phase, but they promise animators more advanced tools to make characters appear more human-like.
One CMU-Disney team is basing facial animations on the best available prototype -- an actual human face.
For now, to animate faces, modern animators must sculpt different emotional expressions and film 3-D models of them. Then the animator blends those expressions together to emulate facial motion.
But Rafael Tena at Disney Research Pittsburgh has produced a method, with help from Carnegie Mellon's Robotics Institute, to capture the motion of an actual actor's face as he or she reads a script and makes dramatic expressions.
By following 13 programmed geometric regions on the face along with a computer program that links certain muscular motions, the new method can generate animated emotions and facial movements that could be far superior than anything currently available.
"Traditionally people who do movies with animated characters use a lot of talented 3-D modelers who sculpt one figure with a mouth open and one with [raised] eyebrows and all the different models with 200 to 300 that can be blended into motion," Mr. Tena said. "We get data from a real actor and define different regions of the face to create a model that allows us to control the face with poses, without having to go through a modeler. Rather than do poses from a sculpture, we develop models from real data."
To make this method of "motion capture" work, the computer analyzes motion in the 13 areas based on which points move in concert and which move independently. The computer program analyzes the subtle motions of the actor's face and produces an animated face replicating them.
First Published September 6, 2011 12:00 am
