The Color Calibration across Multi-Projector Display

Shu Liu, Qiuqi Ruan, Xiaoli LiJournal of Signal and Information Processing 2011 · DOI

Abstract

Large plane displays are often achieved by combining tiled imaging elements, which give the appearance of a single, seamless display. However, this is difficult to realize in practical use, since differences in color will make the tiles distinct even if the geometric calibration is perfect. The paper presents an approach of color calibration using cubic B-spline. According to the theory of cubic B-spline, we construct the color transfer function of each projector. This approach reduces the color and photometric non-uniformities to achieve seamless tiled multi-projector display well. Finally, we propose the difference between blending region and overlapping region which can help to achieve a better effect of alpha blending.

Introduction

Creating expansive, high-resolution displays by tiling multiple projectors is a significant engineering challenge. Even with precise geometric calibration—where pixels are perfectly aligned—displays often suffer from "photometric non-uniformity." This phenomenon causes the tiled output to resemble a patchwork quilt because different projectors, or even different regions within a single projector's field of view, produce varying levels of brightness and color.

While professional-grade instruments like spectroradiometers can address these discrepancies, their high cost makes them inaccessible for many applications. Existing camera-based feedback systems offer a more affordable alternative but frequently struggle with imprecise data. This research bridges the gap by employing cubic B-splines to transform noisy camera feedback into an accurate color correction map.

Methodology

The proposed calibration pipeline involves four primary stages:

  1. Response Sampling: Eighteen sample images per color channel (R, G, and B) are projected at varying intensities and captured using a standard feedback camera.
  2. Cubic B-Spline Modeling: Unlike simple linear interpolation or global polynomial fitting, cubic B-splines are used to model the color transfer function. The "locality" property of B-splines ensures that adjustments to one segment of the curve do not unpredictably alter others, allowing for a smooth and precise representation of the projector's natural "S-curve" response.
  3. Gamut Matching and Linearization: The common dynamic range—the intersection of the minimum and maximum responses across all projectors—is identified. A Look-Up Table (LUT) is then generated to linearize the response, targeting a Gamma of 1.0 to ensure input signals map predictably to light output.
Input-output response curves (S-curves) for different projectors ($F_1, F_2$) and the target linear response ($F$).

Optimized Alpha Blending

To eliminate visible seams in overlapping areas, an alpha-blending function is applied. A key contribution of this work is the distinction between the physical overlapping region and the active blending region.

The transition function $f(x)$ is governed by a power parameter ($p$). As shown in the experimental data, a linear blend ($p=1$) often results in visible edges. Increasing the power parameter creates a steeper, smoother transition in the center of the overlap, which is essential for a seamless visual experience.

The impact of the power parameter $p$ on the curvature of the alpha-blending transition.

Results and Discussion

Empirical testing revealed that the width of the alpha-blending transition should not necessarily match the physical overlap. For a 256-pixel overlap, a blending region of 180 pixels produced the most effective results, eliminating the "bright band" effect common in tiled displays. Furthermore, a power parameter of $p=2$ was found to provide the most seamless transition.

The cubic B-spline approach successfully corrected non-linear projector responses into a linear format, facilitating easier color matching. Qualitatively, the method effectively removed visual artifacts, resulting in a display that appears as a single, continuous image.

Conclusion and Future Work

This study demonstrates a cost-effective method for high-quality color calibration. However, certain limitations remain. The current model assumes an "additive gamut," where R, G, and B channels are independent, ignoring potential channel crosstalk. Additionally, the system's precision is tied to the feedback camera's quality. Future work will explore automated, periodic re-calibration to account for the natural aging and color shifting of projector lamps over time.

Summary updated

Abstract

Large plane displays are often achieved by combining tiled imaging elements, which give the appearance of a single, seamless display. However, this is difficult to realize in practical use, since differences in color will make the tiles dis-tinct even if the geometric calibration is perfect. The paper presents an approach of color calibration using cubic B- spline. According to the theory of cubic B-spline, we construct the color transfer function of each projector. This ap-proach reduces the color and photometric non-uniformities to achieve seamless tiled multi-projector display well. Finally, we propose the difference between blending region and overlapping region which can help to achieve a better effect of alpha blending.
BibTeX
@article{liu2011color,
  author = {Shu Liu and Qiuqi Ruan and Xiaoli Li},
  title = {{The Color Calibration across Multi-Projector Display}},
  journal = {Journal of Signal and Information Processing},
  year = {2011},
  doi = {10.4236/jsip.2011.22008},
}
Figure