[Introduction] [Interface] [Future Work] [Credits] [References]

The Active Panorama project addresses a fundamental problem in videoconferencing, where bandwidth forces a tradeoff between image resolution and field of view. Spherical lenses and mirrors maximizes field of view at the expense of image resolution, and robotics PTZ cameras maximize image resolution at the expense of field of view.  

When we were small we would often look through gift-wrap rolls as if it were an old maritime telescope, inspired by the movies of pirates and ocean adventures. A normal human being has a field of view of about 160 to 208 degrees horizontally; however, most modern cameras can give us only a small fraction of that view (~47 degrees maximum for the Canon VCC4). Hence, lack of context becomes a problem when we apply such systems in videoconferencing, surveillance, and distance learning [6] [2].

Our Active Panorama project provides a context + focus interface for videoconferencing based applications. We use one pre-calibrated pan-tilt-zoom camera to construct a high resolution panoramic image, which serves as context of the remote environment. We superimpose a live video stream on top of the panorama so that the focused activity appear to live in the panorama. We update the background panorama as the camera moves. Since we only use one camera and one live video stream, the system requires very limited bandwidth and hardware support. Foote and Kimber [2] summarize panoramic image generation techniques that are the currently available into the follow table,

We also add our system into this comparison table.

The film-based method stitches multiple high resolution images together to produce one still high resolution image. The quality of the image is very high but it is not live. It is also possible to generate a panorama video with wide-angle lens or fish-eye lens. The problem is that resolution of the image is very poor. A number of researchers have focused on generate a panorama using multiple cameras. The quality of the panorama gets improved but bandwidth requirement gets prohibitive high as the panorama resolution increases. Our approach provide an good tradeoff between the quality of the panorama and bandwidth requirement.

The approach that we are suggesting is to:

1. Capture an image, tagged with the orientation of the camera (current pan/tilt/zoom).
2. Process the image (correct lens distortions and project the image onto a spherical manifold that is characteristic of the camera’s range of motions) according to the position and orientation of the camera.
3. Blend the image into the current panorama based on the orientation coordinates, correcting for the variations in lighting caused by Auto Gain Control (ACG).

To test our approach we have extended Microsoft ® Netmeeting, which is a popular teleconferencing program under the Windows platform, to include a ‘active panorama’ of the conferencing area.

	Generating the Panorama
	The live video superimposed into the panorama
	Selecting viewing frame with box dragging interface
	Camera moves to the desired position The previous frame is updated in the background panoramic image

Back to top

Our current and future work is to provide camera management system [6] to automate the camera operation for applications like distance learning, videoconferencing and survelience.

Merge with sound localization to automate the movement of the camera

Extend with pattern matching methods to create an automatic video surveillance system

Extend with motion tracking so that the system may be used for web lecturing

Back to top