Techniques for Face Motion & Expression Analysis on Monocular Images - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 228

Conclusions and Future Trends
The importance granted to Talking Heads has increased in such a dramatic way
during the last decade, that analysis and synthesis methods developed to
generate face animation are under continuous change to meet the new applica-
tion requirements. Following this trend, the analysis of monocular images to
extract facial motion to be rendered on synthetic 3D-head models has appeared
as a way to simplify and adapt facial animation to current video media. The effort
of this research aims at making facial animation technologies and methods
available to the general public and permitting the study and representation of
already stored image data.
Although the chapter has only covered techniques related to the analysis of face
motion, it is important to remark that there exists a tight relationship between the
methodology used for the analysis and the way the head model is synthesized.
Both analysis and synthesis must share the same semantics and the same syntax
in their motion description. In semantics we include the concept of extracting the
same set of possible actions or analyzed movements that the model can actually
render. By syntax we imply the way this motion is described. Given certain
parameters and magnitudes involved in a specific movement, we should be able
to express and make them represent the same action in the analysis module,
which has generated them, and the synthesis module, which will reproduce them.
Although accomplishing these requirements apparently seems a trivial task, in
current research there is still little implication as to how the motion semantics and
syntax determine the way analysis techniques should be designed. Therefore,
solutions proposed to achieve the same goal, face motion analysis, are rare and
lead to the development of algorithms used only in very specific environments.
There exists a trade-off between the degree of motion detail extracted from the
images and the level of semantic understanding desired. Very precise analysis
techniques that are able to generate information to accurately animate face
models often cannot provide meaningful information about the general facial
expression. Current research trends try to satisfy both needs: accurate motion
analysis and expression understanding, so as to generate better facial motion
synthesis. As a result, the different research perspectives of the scientific
communities involved in the field of facial animation (analysis and synthesis) are
starting to converge.
To be able to synthesize facial motion extracted from media that represent reality
so as to replicate human face behavior in real-time in such a way that we could
no longer distinguish natural from synthetic media can be considered the ultimate
objective of research in facial animation. Every step taken towards this target
allows emerging new technology domains to use Talking Heads in daily/
common-use applications. Telecommunications appears as one of these recent