Hue-based feature detection for geometry calibration of multiprojector arrays

A feature detection method for calibrating multiprojector displays that utilizes a chromatic pattern to avoid the disadvantages of black-and-white features is proposed. Theoretical and experimental analyses implied that our hue-based method was robust to four factors: the ambient light, the non-Lambertian reflection of a display wall, the variation of projectors, and the different positions of cameras. Our method resulted in low rates of misdetection or false detection. Experimental results also indicated that our method was effective to detect weak features. A markerless calibration method was proposed to autoestimate the aspect ratio of the planar display wall and to autocalibrate multiprojectors. Subpixel accuracy was achieved by applying the detected features in the process of geometry calibration of multiprojector displays with embedded processors and cameras.