Easy Calibration of a Projector
We have developed an easy-to-use technique to calibrate a projector. The same technique can also be used to calibrate a projector-camera stereo pair. The technique is useful because the calibration is achieved with a black planar surface (e.g. a wall or screen). No complex patterns or apparatus with Euclidean information are required.
Background & Objective: Several techniques are currently available to calibrate a camera or a stereo pair involving two cameras. For example, camera internal parameters (such as focal length and principal point) can be calculated by viewing a planar checkerboard pattern at two or more different orientations. Relative pose (i.e. external parameters such as translation and rotation) between two cameras in a stereo pair can be calculated if both cameras can see the same checkerboard pattern. However, to date, there have been no simple techniques to calculate internal parameters of a projector. There are two reasons for the difficulty. First, since a projector cannot "see" a pattern, a feedback sensor (e.g. a camera) must be used. Second, for a projector the principal point is intentionally shifted (vertically) from the image center to allow an off-axis projection cone. Thus, the traditional assumption about the principal point (being close to image center) is not valid.
Technical Discussion: We use a rigidly attached camera. The camera is calibrated separately (by observing a printed checkerboard at two or more different orientations). To calibrate the projector, the same attached camera observes checkerboard pattern projected on a blank planar surface, at two or more different orientations with respect to the projector-camera stereo pair. The white planar surface could be a wall or fixed screen, and the rigid stereo pair is rotated to create the different relative orientations. Using two or more homographies between projector and camera, we first compute the parameters up to a projective scale. We then upgrade the parameters in projective coordinates to Euclidean coordinates by assuming a fixed aspect ratio.
Publications:
Technology Area: Computer Vision
Modification Date: July 7, 2008