Color Blending in Outdoor Optical See-through AR: The Effect of Real-world Backgrounds on User Interface Color

Adam Zarger; J. Edward Swan II; Joseph L. Gabbard

REVIEW 4 major objections 5 minor 14 references

Color Blending in Outdoor Optical See-through AR: The Effect of Real-world Backgrounds on User Interface Color

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper finds that color blending between optical see-through AR displays and real-world backgrounds is nearly identical whether the background is a physical material or a painted poster that only metamerically matches it, and that the…

desk verdict A useful optical testbed and a new linear chromaticity-shift observation, but the abstract overstates the poster-proxy evidence and the analysis is qualitative. read the letter →

arxiv 1908.09348 v1 pith:YAAOELAL submitted 2019-08-25 cs.HC

Joseph L. Gabbard , J. Edward Swan II , Adam Zarger This is my paper · ORCID

classification cs.HC

keywords opticalsee-throughaugmentedrealitycolorblendinguserinterfacereal-worldbackgroundscolorimetrychromaticityshiftmetamericmatchadaptiveAR

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that the way optical see-through augmented reality (AR) displays blend interface colors with the real world can be studied using painted posters as stand-ins for actual physical materials. Using a controlled testbed with daylight-quality lighting, the authors overlaid 27 AR colors onto five real backgrounds (brick, foliage, pavement, sand) and four painted posters, measuring the blended light with a colorimeter. They found that real materials and their metameric poster counterparts produce nearly identical color blending, and that the most common failure mode is a linear chromaticity shift in which hues move toward the background color. The authors conclude that UI designers face a compressed color gamut and unpredictable hue shifts, and they propose adaptive systems that predict and compensate for these interactions. The paper is an engineering study; perceptual consequences are hypothesized, not yet verified with human users.

What carries the argument

The load-bearing mechanism is the optical testbed: a D65, 95-CRI light enclosure illuminating real materials or posters, an NVIS nVisor SX oHMD monocle, a 50 mm SLR lens placed at eye position, and an OTC1000 colorimeter that records the blended light in CIE xyY coordinates. The authors convert xyY to CIE 1976 $u'v'$ and $L^*u^*v^*$ space using standard formulas, then compare backgrounds pairwise: for each of the 27 AR colors they draw a vector in $u'v'$ space from the color measured against one background to the color measured against the other, and they decompose the total $L^*u^*v^*$ change into luminance and chromaticity components. This pairwise vector view lets them sort all 55 background pairs into four shift categories: washout by chromaticity, washout by luminance, washout by both, and linear chromaticity shift, plus a fifth category comparing each real material with its poster.

What would settle it

A user-based color-matching experiment in the same testbed: if human observers, adapting to each background, judge the perceived hue and saturation of AR colors in ways that disagree with the colorimeter's predicted shifts — for example, if real materials and their metameric posters produce measurably different perceived colors, or if the predicted linear chromaticity shift does not appear in perception — then the engineering proxy claim would be overturned. A simpler check is to measure the same poster-real pairs under a different light source, such as natural sunlight, and see whether the blended colors still match.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that color-blending behavior in an optical see-through head-mounted display is governed by the chromaticity of the background rather than by its physical composition. When a real brick wall, green foliage, pavement, or sand is replaced by a painted poster of approximately matching color, the measured blended light from 27 overlaid AR colors shifts in almost the same way, despite the poster and the material being only a metameric match. The authors further identify a distinct pattern they call a linear chromaticity shift: when a user moves between two moderately bright backgrounds, all AR colors translate together in CIE 1976 $u'v'$ chromaticity space toward the new background's color, with luminance playing a minor role. Because this shift can turn yellows into oranges, greens into yellows, and blues into purples, it threatens semantically encoded UI colors. The authors present this as an engineering baseline supporting a predictive model for adaptive AR color rendering.

Load-bearing premise

The claim rests on the assumption that the colorimeter, looking through the display optics, reports what a human eye would perceive, and that D65/95CRI lamps faithfully reproduce outdoor sunlight; the paper itself notes that follow-on user studies are needed to verify the conjectured perceptual consequences.

Editorial extensions

If this is right

UI designers for optical see-through displays cannot rely on a fixed color palette; the available gamut compresses toward the background chromaticity, so colors must be chosen with the expected environment in mind.
The linear chromaticity shift means that a user turning from one background to another can experience a systematic repainting of interface hues, a hazard for color-encoded information such as warnings, status, or map features.
Color-matched posters are a valid, cheaper proxy for studying real-world backgrounds in engineering studies of color blending, at least for the materials tested.
An adaptive AR system could measure or predict background chromaticity and dynamically alter rendered UI colors to preserve designer intent.
The finding that luminance contributes little to washout between moderately bright backgrounds suggests that simply increasing display brightness will not fully solve color blending problems.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The metameric-match result hints that the relevant variable is the background's spectral power distribution integrated over the display's additive primaries, so a predictive model could be built from measured background reflectance rather than from material categories.
The linear chromaticity shift category has a natural test in color-constancy experiments: the paper suggests color constancy may explain why this shift has not been reported in user studies, which implies that a human study might find the shift partially corrected by perception, weakening the direct usability claim.
The testbed's methodology could be extended to dynamic backgrounds, such as video see-through displays or moving materials, to see whether the pairwise shift model predicts continuous transitions.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Here's my read. The paper does something genuinely useful: it builds a careful optical testbed that puts real physical materials (brick, foliage, pavement, sand) through an oHMD and measures the blended color with a calibrated colorimeter. That is new work—prior studies used painted posters. The data collection is solid: 31k readings, median aggregation, three white-point calibration checks. And the observation that background changes often produce a linear chromaticity shift in the UI colors—rather than just overall washout—is worth noting; it points at a real hazard for color-coded interfaces.

The main problem is in the packaging. The abstract says the poster/real pairs are 'only a metameric match,' but Section 4 explicitly says there was no attempt to color-match the posters to the physical materials. Only brick and pavement were close; green foliage is called out as not close, and brown foliage and sand have no poster counterpart. So the poster-proxy equivalence claim is built on three pairs, two of which were near-matches, and the green foliage comparison shows a linear shift, not equivalence. That is not evidence for metameric equivalence; it is evidence that a couple of custom posters came close to their materials. The abstract needs to be rewritten.

The analysis is also thinner than I'd like. The five washout/shift categories were hand-sorted from printed small multiples. The authors admit categories (b) and (c) are not clearly separated. There are no error bars, no confidence intervals, no inferential statistics. For an engineering study that is not necessarily fatal—the raw measurements are objective—but it means the headline 'nearly the same' is more of a visual impression than a measured result.

Two more limitations, both acknowledged by the authors: the 95CRI fluorescent lamps are a good but imperfect proxy for daylight, and the colorimeter looking through an SLR lens is not a human eye. The paper explicitly defers user studies for the perceptual claims. I don't hold that against it, but it does mean the title's 'Outdoor' context and phrases like 'crippling the usability of the UI' are extrapolations, not findings.

Who should read this? Researchers who build AR display testbeds and anyone studying color in optical see-through UIs. It deserves peer review—the testbed and the linear-shift observation are worth exposing to a community that is still mostly guessing about color—but it needs revision: correct the abstract, add error bars and at least basic statistical comparisons, and make the data available. I'd accept a revised version, but not as-is.

Referee Report

4 major / 5 minor

Summary. This paper reports an engineering measurement study of color blending in outdoor optical see-through AR. The authors constructed an optical testbed with 95 CRI fluorescent lamps intended to reproduce D65 daylight, an NVIS nVisor SX oHMD monocle, an SLR lens, and a colorimeter. They measured 27 oHMD display colors against 11 background conditions (five physical materials, four painted posters, a white poster, and a no-lights condition), collecting 31,195 readings and reducing each background-by-color cell to a median xyY value, which they transformed into u'v' and L*u*v*. The analysis is a qualitative pairwise comparison: 55 small-multiple plots were hand-sorted into four categories of color shift plus a poster-versus-real category. The paper claims that physical material backgrounds blend with oHMD colors nearly the same as their corresponding poster backgrounds, and that a newly identified linear chromaticity shift creates a usability hazard for color-coded UIs.

Significance. If the empirical results are accepted, the paper provides a useful testbed design and a substantial baseline dataset (27 colors x 11 backgrounds, with objective colorimeter measurements) for oHMD color research. The L*u*v* decomposition into luminance and chromaticity components is a reasonable way to characterize color shifts. The paper's strengths include systematic stimulus generation, explicit calibration against D65 white points, and a clear comparison between physical materials and painted posters. However, the central claims about 'metameric' poster-real equivalence and the 'crippling' usability consequences go beyond the instrument-based data; the qualitative hand-sorted analysis lacks statistical support, and the paper itself defers perceptual validation to future user studies. The contribution is therefore best viewed as an engineering measurement study that generates hypotheses, not as the final word on outdoor UI color perception.

major comments (4)

[Abstract; Section 4; Section 5] The abstract and Section 5 state that the poster and physical background pairs are 'only a metameric match,' but Section 4 states 'there was no attempt to color-match poster backgrounds with their physical material counterparts; although in two cases the matching was very close (brick and pavement).' Metamerism is a spectral property that requires colorimetric equivalence under a given illuminant, and no spectral or colorimetric matching was performed for most pairs. This contradiction undermines the abstract's premise for the poster-proxy comparison, although it does not invalidate the raw measurements.
[Section 5; Figure 5] The five categories in Figure 5 are produced by hand-sorting printed small multiples; no error bars, confidence intervals, or statistical tests are reported, and the paper concedes that categories (b) and (c) are not clearly separable. Because the novel category (d) ('linear chromaticity shift') and the poster-real equivalence category (e) drive the main conclusions, the analysis should be supported by quantitative criteria (for example, thresholds on vector direction or length in u'v' and L*u*v*) or by inter-rater reliability checks. Additionally, only three poster-real pairs are compared (brick, green foliage, pavement); brown foliage and sand have no poster counterpart, so the claim that color blending effects are 'nearly the same' rests on a small, non-random sample.
[Section 5; Section 6] The usability conclusions in the abstract and Section 5—'significantly crippling the usability of the UI,' altered semantic interpretation, and specific hue shifts such as 'yellows shift into oranges, greens shift into yellows, and blues shift into purples'—are inferred from colorimeter measurements rather than from human perception. The paper itself states 'human judgments are required' and 'we need to conduct follow-on user studies to verify the perceptual consequences we conjecture above' (Section 5). These statements are internally consistent but the abstract overstates them as established findings; the conclusions should be reframed as predictions to be tested in user studies.
[Section 3; Section 5] The testbed's 95 CRI lamps are described as reproducing the D65 daylight standard white point, but CRI 95 does not guarantee a spectral match sufficient for metamerism claims under the experimental geometry, and the paper does not report spectral measurements of the light source or the reflected backgrounds. Because the title and abstract frame the study as 'outdoor' optical see-through AR, the authors should either provide evidence that the lighting approximation is adequate for the claims or explicitly scope the conclusions to the testbed's simulated conditions.

minor comments (5)

[Section 5] The heading 'Poser vs. Real Comparisons' contains a typo; it should read 'Poster vs. Real Comparisons.'
[Author bios] The second author biography says 'formally with Virginia Tech'; this should be 'formerly with Virginia Tech.'
[Section 1] There are typos in the introduction: 'AR attempts to is affirm its position' should be 'AR attempts to affirm its position,' and 'effected' in the first sentence should be 'affected.'
[Section 4 footnote] The footnote calculating N = 296 is typeset awkwardly; it should be written clearly as 10 backgrounds x 27 colors + 1 background x 26 colors = 296 cells, with an explanation that the no-lights condition cannot measure the black (oHMD-off) color.
[Section 5; Figure 5] The small-multiple analysis is based on printed graphs that are not reproducible from the manuscript alone; the paper would benefit from a data availability statement or supplementary data files so that readers can verify the hand-sorted categories.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the poster/real comparison is direct measurement, not a fitted or self-cited prediction.

full rationale

This is an empirical engineering measurement study, not a derivation or prediction chain. The paper reports direct colorimeter measurements of blended light under three lighting conditions for 27 display colors and 11 backgrounds (Section 4), reducing the raw readings to median xyY values and then applying standard CIE color-space formulas. No model parameter is fitted to a subset of the data and then used to 'predict' a closely related quantity; the poster-versus-physical-material comparison is made from measurements of both kinds of backgrounds taken in the same testbed under the same apparatus. The citation to the authors' prior work [1] is used for the enclosure lighting design and for the earlier poster-based methodology, but the current paper's own data include both poster and physical material conditions, so the central equivalence claim does not reduce to a self-citation. The statement in Section 5 that 'we need to conduct follow-on user studies to verify the perceptual consequences we conjecture above' is an acknowledged validity limitation about generalizing from colorimeter readings to human perception, not a circular dependency. The abstract's 'metameric match' phrase is in tension with Section 4's statement that 'there was no attempt to color-match poster backgrounds with their physical material counterparts,' but that is an internal consistency or evidentiary concern, not a circularity. No equation in the paper is defined in terms of the conclusion being drawn, and the 'linear chromaticity shift' category is an empirical observation from measured u'v' scatterplots rather than a result derived from the categories themselves. Therefore the paper is self-contained with respect to circularity concerns and warrants a score of 0.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper is an empirical measurement study with no fitted parameters or new theoretical entities. Its central claims rest on the validity of the testbed optics and lighting, the choice of representative materials, and the appropriateness of the colorimetric transformations.

assumptions (4)

domain assumption The colorimeter and SLR lens assembly measures the same light that would enter a user's eye through the oHMD monocle.
The testbed places a 50 mm SLR lens and colorimeter at the approximate eye position (Section 3), but no user validation is reported; Section 5 states that user studies are needed to verify perceptual consequences.
domain assumption Fluorescent lamps with 95 CRI and D65 white point adequately emulate outdoor daylight for color blending purposes.
The testbed uses 95 CRI fluorescent lights to represent natural outdoor illumination (Section 3); real sunlight has a different spectral power distribution, and the paper does not compare against actual outdoor light.
ad hoc to paper The median xyY value for each background-by-color cell is a valid central tendency for the measured color.
Section 4 describes aggregating 1,459 to 7,595 readings per cell into a single median, but no variance or time-stability analysis is reported; the paper does not justify median over mean.
standard math Standard CIE colorimetry equations (xyY to XYZ to u'v' to L*u*v*) apply as given in Foley et al.
Section 4 uses equations (1)-(4) from Foley et al. [17, 18] to transform measured xyY values; these are established standards.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Color Blending in Outdoor Optical See-through AR: The Effect of Real-world Backgrounds on User Interface Color." pith.science (2026). https://pith.science/paper/YAAOELAL

@misc{pith2026190809348,
  author       = {Pith},
  title        = {Pith review of: Color Blending in Outdoor Optical See-through AR: The Effect of Real-world Backgrounds on User Interface Color},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YAAOELAL}},
  note         = {Machine review of arXiv:1908.09348}
}

read the original abstract

It has been noted anecdotally and through a small number of formal studies that ambient lighting conditions and dynamic real-world backgrounds affect the usability of optical see-through augmented reality (AR) displays; especially so in outdoor environments. Our previous work examined these effects using painted posters as representative real-world backgrounds. In this paper, we present a study that employs an experimental testbed that allows AR graphics to be overlaid onto real-world backgrounds as well as painted posters. Our results indicate that color blending effects of physical materials as backgrounds are nearly the same as their corresponding poster backgrounds, even though the colors of each pair are only a metameric match. More importantly, our results suggest that given the current capabilities of optical see-through head-mounted displays (oHMDs), the implications are, at a minimum, a reduced color gamut available to user interface (UI) designers. In worse cases, there are unknown or unexpected color interactions that no UI or system designers can plan for; significantly crippling the usability of the UI or altering the semantic interpretation of graphical elements. Further, our results support the concept of an adaptive AR system which can dynamically alter the color of UI elements based on predicted background color interactions. These interactions can be studied and predicted through methods such as those presented in this work.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages

[1]

transforms xyY into 1931 XYZ CIE primaries, (eq

work page 1931
[2]

transforms XYZ into 1976 CIE u’v’ values, (eq

work page 1976
[3]

also representing the color of typical real-world backgrounds: brick poster, green foliage poster, pavement poster and sidewalk poster. By using these posters, we were able to make interesting 1 There were 10 (background: brick poster, real brick, real brown foliage, green foliage poster, real green foliage, pavement poster, real pavement, real sand, side...

work page 1931
[4]

5 RESULTS AND DISCUSSION To analyze the data, we created a small-multiples graph that de-picted all 55 pairwise combinations of our 11 backgrounds

transforms Y, Yn, u’v’, and un’vn’ into L*u*v* values. 5 RESULTS AND DISCUSSION To analyze the data, we created a small-multiples graph that de-picted all 55 pairwise combinations of our 11 backgrounds. Our motivation for looking at pairwise combinations is the idea that an AR user would be facing one background, and then turn to face another background; ...

work page 1976
[6]

More Than Meets the Eye: An Engineering Study to Empirically Examine the Blending of Real and Virtual Color Spaces,

J. L. Gabbard, J. Zedlitz, J. E. Swan II, and W. W. Winchester, "More Than Meets the Eye: An Engineering Study to Empirically Examine the Blending of Real and Virtual Color Spaces," in IEEE conference on Virtual Reality, Waltham, MA, 2010, pp. 79-86

work page 2010
[7]

MR in OR: First analysis of AR/VR visualization in 100 intra-op-erative Freehand SPECT acquisitions,

A. Okur, S.-A. Ahmadi, A. Bigdelou, T. Wendler, and N. Navab, "MR in OR: First analysis of AR/VR visualization in 100 intra-op-erative Freehand SPECT acquisitions," in Mixed and Augmented Reality (ISMAR), 2011 10th IEEE International Symposium on, 2011, pp. 211-218

work page 2011
[8]

Evolutionary augmented reality at the Natural History Museum,

P. Debenham, G. Thomas, and J. Trout, "Evolutionary augmented reality at the Natural History Museum," in Mixed and Augmented Reality (ISMAR), 2011 10th IEEE International Symposium on, 2011, pp. 249-250

work page 2011
[9]

Augmented Reality: An Applica-tion of Heads-Up Display Technology to Manual Manufacturing Processes,

T. P. Caudell and D. W. Mizell, "Augmented Reality: An Applica-tion of Heads-Up Display Technology to Manual Manufacturing Processes," in 25th Hawaii International Conference on System Sci-ences (HICSS), 1992, pp. 659–669

work page 1992

Show all 14 references

[10]

The Argon AR Web Browser and standards-based AR application envi-ronment,

B. MacIntyre, A. Hill, H. Rouzati, M. Gandy, and B. Davidson, "The Argon AR Web Browser and standards-based AR application envi-ronment," in Mixed and Augmented Reality (ISMAR), 2011 10th IEEE International Symposium on, 2011, pp. 65-74

2011
[11]

Visualiza-tion Techniques for Augmented Reality Handbook of Augmented Reality,

D. Kalkofen, C. Sandor, S. White, and D. Schmalstieg, "Visualiza-tion Techniques for Augmented Reality Handbook of Augmented Reality," B. Furht, Ed., ed: Springer New York, 2011, pp. 65-98

2011
[12]

Active Text Drawing Styles for Outdoor Augmented Reality: A User-Based Study and Design Implications,

J. L. Gabbard, J. E. Swan II, D. Hix, S. J. Kim, and G. Fitch, "Active Text Drawing Styles for Outdoor Augmented Reality: A User-Based Study and Design Implications," in IEEE conference on Virtual Re-ality, Charlotte, North Carolina, USA, 2007, pp. 35-42

2007
[13]

Quantification of Contrast Sensitivity and Color Perception using Head-worn Aug-mented Reality Displays,

M. A. Livingston, J. H. Barrow, and C. M. Sibley, "Quantification of Contrast Sensitivity and Color Perception using Head-worn Aug-mented Reality Displays," presented at the IEEE Virtual Reality 2009, Lafayette, LA, USA,

2009
[14]

Assessing the Usability of a Wearable Computer for Outdoor Pedestrian Navigation,

T. Pingel, and Clarke, K.C., "Assessing the Usability of a Wearable Computer for Outdoor Pedestrian Navigation," presented at the Au-toCarto 2005, Las Vegas, NV,

2005
[16]

These medians reduced our collected dataset to one value per back-ground by color combination, giving N = 296 data points1

for details). These medians reduced our collected dataset to one value per back-ground by color combination, giving N = 296 data points1. Lastly, we transformed the data into the CIE 1976 u’v’ and L*u*v* color spaces, using the formulas found in Foley et al. [17]. There is a d...

1976

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

transforms xyY into 1931 XYZ CIE primaries, (eq

work page 1931

[2] [2]

transforms XYZ into 1976 CIE u’v’ values, (eq

work page 1976

[3] [3]

also representing the color of typical real-world backgrounds: brick poster, green foliage poster, pavement poster and sidewalk poster. By using these posters, we were able to make interesting 1 There were 10 (background: brick poster, real brick, real brown foliage, green foliage poster, real green foliage, pavement poster, real pavement, real sand, side...

work page 1931

[4] [4]

5 RESULTS AND DISCUSSION To analyze the data, we created a small-multiples graph that de-picted all 55 pairwise combinations of our 11 backgrounds

transforms Y, Yn, u’v’, and un’vn’ into L*u*v* values. 5 RESULTS AND DISCUSSION To analyze the data, we created a small-multiples graph that de-picted all 55 pairwise combinations of our 11 backgrounds. Our motivation for looking at pairwise combinations is the idea that an AR user would be facing one background, and then turn to face another background; ...

work page 1976

[5] [6]

More Than Meets the Eye: An Engineering Study to Empirically Examine the Blending of Real and Virtual Color Spaces,

J. L. Gabbard, J. Zedlitz, J. E. Swan II, and W. W. Winchester, "More Than Meets the Eye: An Engineering Study to Empirically Examine the Blending of Real and Virtual Color Spaces," in IEEE conference on Virtual Reality, Waltham, MA, 2010, pp. 79-86

work page 2010

[6] [7]

MR in OR: First analysis of AR/VR visualization in 100 intra-op-erative Freehand SPECT acquisitions,

A. Okur, S.-A. Ahmadi, A. Bigdelou, T. Wendler, and N. Navab, "MR in OR: First analysis of AR/VR visualization in 100 intra-op-erative Freehand SPECT acquisitions," in Mixed and Augmented Reality (ISMAR), 2011 10th IEEE International Symposium on, 2011, pp. 211-218

work page 2011

[7] [8]

Evolutionary augmented reality at the Natural History Museum,

P. Debenham, G. Thomas, and J. Trout, "Evolutionary augmented reality at the Natural History Museum," in Mixed and Augmented Reality (ISMAR), 2011 10th IEEE International Symposium on, 2011, pp. 249-250

work page 2011

[8] [9]

Augmented Reality: An Applica-tion of Heads-Up Display Technology to Manual Manufacturing Processes,

T. P. Caudell and D. W. Mizell, "Augmented Reality: An Applica-tion of Heads-Up Display Technology to Manual Manufacturing Processes," in 25th Hawaii International Conference on System Sci-ences (HICSS), 1992, pp. 659–669

work page 1992