The auto-geometric correction of multi-projector for cylindrical surface using Bézier patches

Geometric calibration to projection images is an indispensable operation for projection-based spatial display. In this paper, we propose a new method for correcting images generated in a computer onto a cylindrical surface accurately, which can project a high-resolution projection image with pixels matching avoiding too much manual operation. Images waiting to be projected are pre-warped according to the rough correspondence between projectors and physical surface. To solve the errors resulting from unexpected pixel shifts in overlap projection area, we fit the Bézier interpolation to the images and apply the optimization theory with added constraints to correct the projection image accurately. This optimization process, by taking the pixels with specific significance on the images as the basis of calculation, avoids the traditional ways of translating the control points of the Bézier surface directly. The final results achieve a completely accurate projection picture even if the projection surface shape is inaccurate and irregular. We present the details of the proposed accurate calibration algorithm and illustrate our method, which, with its scalability, can achieve perfect projection efficiently and accurately with experiments.