Computer vision based geometric calibration in curved multi-projector displays

In a multi-projector display system, geometric calibration is usually time-consuming and tedious. In this paper, we propose a multi-projector tiled display method with geometry parameters estimation. The method is used for analytic curved surface, such as cylindrical or spherical shaped display. The calibration process is automatic by using computer vision techniques; and in order to estimate the screen and projector parameters, a camera stereo pair with given internals is used to reconstruct 3D points on the screen. In this parametric approach, the camera doesn't have to cover the entire screen and the projectors can be placed roughly, which give the flexibility for various multi-projector display system. For scene rendering, our implementation is distributed, which is more efficient than the image transfer method with modern GPU. At last, we analyzed the effectiveness of the proposed method. Experimental results show that our approach supplied sufficient accuracy and enhanced the immersiveness.