CONCLUSION AND FUTURE WORK - 3D Face Processing: Modeling, Analysis and Synthesis - page 108

Chapter 10
CONCLUSION AND FUTURE WORK
In this topic, we have presented a unified 3D face processing framework.
Various aspects of face processing research have been discussed in the context
of multi-modal human computer interaction and intelligent video analysis. In
this chapter, we summarize the contributions of this topic and outline the future
research directions.
1. Conclusion
In this topic, we have presented a unified 3D face processing framework.
In Chapter 2, we describe tools for building 3D geometric models of neutral
faces. Using these tools, we can create personalized 3D face models for 3D face
tracking and animation. Then we discuss the geometric facial motion models
in the framework in Chapter 3. These motion models are derived from motion
capture data of real face motions. Thus they can capture the characteristics
of real face motions. After that, we present our approaches for 3D non-rigid
face tracking and animation in Chapter 4 and Chapter 5, respectively. We
demonstrate the efficacy of tracking and animation, using experimental results
in very low bit-rate face video coding, speech-drive face animation, and etc.
In Chapter 6, we present flexible appearance model to deal with appearance
details which are lost in the geometric facial motion models. In out experiments,
these details are shown to be important for computer analysis of subtle facial
expressions [Wen and Huang, 2003], and human perception of synthesized
face animations [Wen et al., 2003]. Compared to most existing appearance
models for face motions, our flexible appearance model is less illumination
dependent, and less person-dependent. It also requires less data for estimating
the parameters of the flexible appearance model. Therefore, our appearance
model can be more flexibly used in various environments.
115