Automatic 3D Face Model Adaptation with Two Complexity Modes for Visual Communication - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 303

parameters ( d u , d o ) stand for the thickness of the lips. The mouth width L is
calculated as the distance between the left mouth corner point h l and the right
mouth corner point h r . To represent the mouth-open features, six parameters
are needed: (i) two mouth corner points, (ii) two thickness of the lips, and (iii) two
opening heights of the lips. For the representation of the mouth-closed features,
five parameters are needed: (i) two mouth corner points, (ii) two thickness of the
lips, and (iii) one parameter t which refers to the height between the level of the
corners of the mouth and the contact point of the two lips.
Based on the representation of eye and mouth features using the 2D parametric
models described above, the parameters of the models are separately estimated
one after another (ref. Figure 5). The search areas for eye and mouth features
are first determined using the algorithm proposed in Kampmann & Ostermann
(1997). Within these search areas, the pupil points and the corner points of the
eyes and the mouth are estimated with template matching techniques (Zhang,
1997). After that, these estimated points are utilized to fix the search areas for
the lip thickness and the opening heights of the lips and the eyelids. Since two
mouth models are exploited for representing the 2D mouth features, an appro-
priate mouth model has to be automatically selected first. The lip thickness and
the opening heights of the lips will then be estimated by minimization of cost
functions using an appropriate 2D mouth model. The eyelid opening heights are
also estimated using the 2D eye model analogous to the estimation of the lip
Figure 5. Flowchart of eye and mouth feature estimation.