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 307

occurrence of the chin and cheek contours is calculated for each pixel. This
calculation takes advantage of the fact that the chin and cheek contours are more
likely to be located in the middle of the search area than at the borders of the
search area. An estimation rule is established using this probability and assumes
high image gradient values at the position of the chin and cheek contours. By
maximization, the position of the chin contour (ref. Figure 7a), as well as the
positions of the cheek contours (ref. Figure 7b) are estimated.

Eyebrows

For estimation of eyebrow features, some knowledge about these features is
exploited (Kampmann & Zhang, 1998):

1. Eyebrows are located above the eyes.

2. Eyebrows are darker than the surrounding skin.

3. Eyebrows have a typical curvature, length and thickness.

4. Eyebrows could be covered by hair.

Using this knowledge, eyebrow features are estimated. First, using the estimated
eye positions, search areas for the eyebrows are introduced above the eyes.
Then, a binarization of the luminance image using a threshold is carried out. In
order to determine this threshold, the upper edge of the eyebrow is identified as
the maximum value of the luminance gradient. The threshold is now the mean
value between the luminance value of the skin and the luminance value of the
eyebrow at this upper edge. After binarization, the area with a luminance value
below the threshold is checked whether it has the typical curvature, length and
thickness of an eyebrow. If the answer is yes, the eyebrow is detected. If the
answer is no, the eyebrow is covered by hair. For this case, using the
morphological image processing operations erosion and dilation, as well as
knowledge about the typical shape of an eyebrow, it is decided whether the
eyebrow is completely or only partly covered by hair. In the case that the

Figure 8. Estimation of eyebrow features: (a)(b)(c) Original image, (d)(e)(f)
Eyebrows after binarization, (g)(h)(i) Eyebrows after removal of hair.

(a) (b) (c) (d) (e) (f) (g) (h) (i)