Adaptative projection on a mobile planar surface

A similar solution has already been proposed by [1]. The authors have developed a system able to project video content on a stationary planar surface while moving a projector. To continuously adapt the projection, they rely on the three homographies between the screen, the camera and the projector. Since these homographies are related and the one between the screen and the camera does not change over time, the only homography that needs to be updated is the one linking the projector and the camera (the one between the projector and the screen is indeed impossible to compute directly). To compute this homography, they use correspondences between feature points found on the projector and camera images. These feature points are extracted from the content thanks to the SIFT method and are