While the traditional approach to collecting video data for stereoscopic display is to use two cameras, it suffers from several drawbacks including excess cost and size, added complexity, and the necessity of careful calibration. Within Phase I, we investigated techniques for collecting stereoscopic data from a single camera, which eliminated the problems associated with the dual-camera configuration. In the Phase I effort, we used Equinox's ARC device to perform this real-time stereoscopic display task while at the same time doing so within a small, low-power, low-cost device. While in Phase I, the device was limited to very simple rectification algorithms to correct for camera rotations, we did investigate more advanced rectification algorithms within a software system running on a PC workstation. In Phase II, we propose to modify the existing ARC device to produce a custom architecture for stereoscopic display and image correction. By leveraging our in house technology, we can do so rapidly and at low cost. The custom architecture will allow more advanced rectification algorithms to be deployed, while at the same time keeping the physical and power footprint limited to something not significantly more than a thicker iPod.
Keywords: Time-Sequential Stereo, Stereoscopic 3d Video, Surveillance
