Parameterized Models for Facial Animation
There are several approaches to applying computer graphics techniques to
Both approaches have some drawbacks, here are some suggested by K. Waters.
- Key frame animation.
The desired image is specified for a certain point in time and then again
for another point in time a certain number of frames later. A computer
algorithm generates the necessary images for the frames in between these
This technique has been successfully adapted to 2D and 3D animation.
- Parameterized models.
Two basic idease are involved in the creation of parametrized graphical models:
- determine the appropriate set of parameters, and
- develop image synthesis models based on these parameters.
Developing parameter sets
How is a facial parameter set developed?
Several approaches are possible:
There are two broad categories of parameters; those controlling the
- Observe the surface properties of faces and develop ad hoc sets that allow
these observed characteristics to be specified parametrically.
- Studying the underlying structure, or facial anatomy, and developing a set
of parameters based on it -- the approach used by Platt and Badler.
- Parameters are based on structural understanding, wherever possible, and are
supplemented as necessary by parameters based on observation -- the approach
used by Parke.
What should the parameters for an expression model be? A possible
foundation -- and it appears to be the best basis -- is the
Facial Action Coding System, or FACS, developed by Paul Ekman.
- conformation, or structure, of an
- expression, or emotional content.
Which parameters are best included in a simple expression model remains
unresolved. Of primary importance in expression are the eyes and mouth,
so most parameters relate to these areas.
For the eyes:
For the mouth:
- pupil dilation
- eyelid opening
- eyebrow position and shape
- the direction in which the eyes are looking
- jaw rotation (controls mouth opening)
- width of the mouth
- mouth expression (smiling, frowning, ...)
- position of the upper lip
- positions of the corner of the mouth
- size of the nostrils
- the orientation of the face
Changes in the confirmation of faces (those aspects that vary from individual
to individual and make each perso unique) require a different set of parameters.
Again, the ideal set is unknown.
- control the color of the eyebrows, eyelashes, iris, lips, ...
- use relative size (scale), shape and positioning information to control:
- neck length and shape
- chin, forehead, cheek, and cheekbone shape
- eyelid, eyeball, and iris size and the position and separation of the eyes
- jam width
- nose length and the width of the bridge and end of the nose
- chin and forehead scale and the scale vis-a-vis the rest of the face.
The development of truly complete confirmation parameters appears very difficult.
One such parametrized model was given by K Waters
Back to home..