Home	Concept	Technology	Budget	About
Image Construction - Tracking and Rendering - Challenges One significant issue with this system is occlusion. When two or more people are in the space, it may be impossible to reconstruct an adequate image of the head. In the extreme case, one person may be surround by many others resulting in no wall-camera information regarding that person. However, since the overhead camera records the absolution position of each person in the room, it is possible to compute how much useful data will be returned by each wall-mounted camera. Since the aesthetics of the project indicate that only about five people should be tracked at once (resulting in 15 virtual heads), the simplest solution is simply to determine which five individuals are more trackable and always work with these. The individuals who are tracked will likely change in a natural way as people move around the space. Another issue is real-time display. The system described displays a static image of the head recorded at the moment the person enters the room. A more powerful system would allow the 3D reconstruction to take place in real-time, permitting real-time reconstructions to be rendered. Although five real-time camera streams will enter the server in parallel, the factor in this case is likely computing power and not bandwidth as these camera streams must be back-projected in real-time. A simplifaction would transmit only the center of the face so that important features (like the eyes and mount) are real-time, while the rest of the head would be static (which it is normally). Note that static in this case refers to the skin features themselves, but not to orientation and position which would change dynamically even in the system described above.