We present a novel calibration framework for multi-projector displays that achieves continuous geometric calibration by estimating and refining the poses of all projectors in an ongoing fashion during actual display use. Our framework provides scalability by operating as a distributed system of "intelligent" projector units: projectors augmented with rigidly-mounted cameras, and paired with dedicated computers. Each unit interacts asynchronously with its peers, leveraging their combined computational power to cooperatively estimate the poses of all of the projectors. In cases where the projection surface is static, our system is able to continuously refine all of the projector poses, even when they change simultaneously.

Using cameras to geometrically calibrate projector-based displays has been widely reported in the literature over the last decade. Most systems project structured-light patterns during a setup phase before starting the application in order to evaluate the geometric mapping from projector pixels to locations on the display surface. If this mapping changes e.g. the projector is moved, the application must be stopped and calibration re-initiated. Our pose estimation technique is based on detecting a set of image correspondences between the moving projector and a static camera using only the imagery displayed by the projector. We use an up-front calibration process to establish the geometry of the display surface, the pose of the camera, and the initial pose of the projector.

Projectors equipped with wide-angle lenses can have an advantage over traditional projectors in creating immersive display environments since they can be placed very close to the display surface to reduce user shadowing issues while still producing large images. However, wide-angle projectors exhibit severe image distortion requiring the image generator to correctively pre-distort the output image. In this paper, we describe a new technique based on Raskar's (1998) two-pass rendering algorithm that is able to correct for both arbitrary display surface geometry and the extreme lens distortion caused by fisheye lenses. We further detail how the distortion correction algorithm can be implemented in a real-time shader program running on a commodity GPU to create low-cost, personal surround environments

Using projectors to create perspectively correct imagery on arbitrary display surfaces requires geometric knowledge of the display surface shape, the projector calibration, and the user’s position in a common coordinate system. Prior solutions have most commonly modeled the display surface as a tessellated mesh derived from the 3D-point cloud acquired during system calibration. In this paper we describe a method for functional reconstruction of the display surface, which takes advantage of the knowledge that most interior display spaces (e.g. walls, floors, ceilings, building columns) are piecewise planar. Using a RANSAC algorithm to recursively fit planes to a 3D-point cloud sampling of the surface, followed by a conversion of the plane definitions into simple planar polygon descriptions, we are able to create a geometric model which is less complex than a dense tessellated mesh and offers a simple method for accurately modeling the corners of rooms. Planar models also eliminate subtle, but irritating, texture distortion often seen in tessellated mesh approximations to planar surfaces.