Graphics Reference
In-Depth Information
Chapter 6
FLEXIBLE APPEARANCE MODEL
In previous chapters we have presented the geometric facial motion model‚
and how the model can be used in facial motion analysis and synthesis. Geomet-
ric motion model handles the macro- and meso-structure level deformations.
However‚ facial motions also exhibit detailed appearance changes such as wrin-
kles and creases as well. These details are important visual cues for both human
perception and computer analysis. Nevertheless‚ they are difficult to analyze
and synthesize using geometric motion models. In this chapter‚ we introduce
our flexible appearance model in our 3D face processing framework. It aims to
deal with facial motion details‚ thus enhance the geometric model based anal-
ysis and synthesis. However‚ the space of all face appearance is huge‚ affected
by the variations across different head poses‚ individuals‚ lighting‚ expressions‚
speech and etc. Thus it is difficult for most existing appearance-based methods
to collect enough face appearance data and train a model that works robustly
in many different scenarios. Compared to these approacnes‚ the novelty of our
flexible appearance model is that we propose efficient and effective methods
to reduce illumination dependency and person dependency of the appearance
model‚ based on limited data. Therefore‚ it is more flexible to use the appearance
model in varying conditions.
In Section 1‚ we first give a review of the related works on utilizing appearance
models for facial motion analysis and synthesis. Then we introduce our flexible
appearance model in Section 2. Section 2.1 describes the illumination modeling
component of the flexible appearance model. Then‚ we discuss the technique
for reducing person dependency in Section 2.2. Analysis and synthesis using
the flexible model will be presented later in Chapter 7 and 8‚ respectively.
Search WWH ::
Custom Search