next up previous contents
Next: Synthesizing Segmented Mug-Shot Images Up: Face Normalization and Recognition Previous: Mirroring Along the Mid-Line

Synthesizing 2D Images with a Texture Mapped 3D Model

Having fully coated the 3D model with the intensity image data, we can now rotate, translate and stretch the model arbitrarily to synthesize new views of the individual in question. By selecting a new set of parameters $(t_{x},t_{y},t_{z},\theta_{x},\theta_{y},\theta_{z},s_{y})$ we can reposition our model and re-project a new 2D image of the original face. Sample re-projections of a 2D face obtained from a single view image are shown in Figure [*].


  
Figure 4.15: Some re-projections of the coated 3D model.
\begin{figure}\center
\begin{tabular}[b]{ccc}
\epsfig{file=norm/figs/fordi1a....
...,height=4cm}\\
(a) & (b) & (c)
\end{tabular} \\ \vspace*{0.5cm}
\end{figure}



 

Tony Jebara
2000-06-23