The authors describe a camera-based tracking system for immersive displays where users are interacting with a virtual environment in front of a projection screen.
The motion tracking technique proposed involves four important components: an image synchronization step, a geometric matching transformation, and a color transformation step, followed by the actual segmentation procedure. The image synchronization step triggers the frame capture at specific time intervals, assuring a correct frame capture. The geometric matching transformation is a calibration procedure that aligns the captured frame with the projection screen. The color transformation compensates for the color distortions due to image projection. Finally, the segmentation procedure distinguishes the background from the foreground pixels.
The segmentation procedure is followed by an optional skeleton fitting procedure. The skeleton fitting is a variation of existing methods, and has a cycle time of about 27 milliseconds. The authors fail to mention the combined times of the previous procedures. For the tracking results to be useful in a real-time virtual reality application, assuming a frame rate of 30 frames per second, the cycle time of all of the processing steps combined should be less than 33 milliseconds.
The methods presented in this paper provide a potential approach for markerless motion tracking in virtual environments. Aside from a few difficult sentences, the text is easily readable and the methods are described at a fair level of detail, with pointers to helpful additional references.