We will develop novel algorithms and hardware to support multi-projector display that can display any content on anywhere at anytime. Key to the innovations is to use computer vision algorithms to provide closed-loop control. That is, a camera is used to observe the setup and automatically estimate (calibrate) the necessary adjustments needed to register the imagery, in terms of both geometry and color. While others have shown success using this camera-based approach, unique to this proposal is to calibrate these parameters continuously while the display system is in use. This autocalibration capability was first developed by Dr. Yang (the research partner for this proposal) and Dr. Welch in 2001. In this proposal, we will extend the original approach to include all parameters for display calibration, allowing both display surface deformation and projector movement. By maintaining an optimal calibration all the time, the display is more robust, can withstand rough handling, and can be modified on the fly. We will also demonstrate, at no cost to the funding agency, a visual training software (or any software), without any modification, running with our multi-projector display. This application-transparency, together with autocalibration, will greatly improve the usability of multi-projector displays.
Keywords: Large-Scale Displays, Autocalibration, Imaging Geometry, Kalman Filter, Graphics Systems