Content uploaded by Trevor Canham
Author content
All content in this area was uploaded by Trevor Canham on Apr 17, 2019
Content may be subject to copyright.
The authors are solely responsible for the content of this technical presentation. The technical presentation does not necessarily reflect the
official position of the Society of Motion Picture and Television Engineers (SMPTE), and its printing and distribution does not constitute an
endorsement of views which may be expressed. This technical presentation is subject to a formal peer-review process by the SMPTE
Board of Editors, upon completion of the conference. Citation of this work should state that it is a SMPTE meeting paper. EXAMPLE:
Author's Last Name, Initials. 2018. Title of Presentation, Meeting name and location.: SMPTE. For information about securing permission
to reprint or reproduce a technical presentation, please contact SMPTE at jwelch@smpte.org or 914-761-1100 (445 Hamilton Ave., White
Plains, NY 10601).
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®)
Influence of Ambient Chromaticity on Portable Display
Color Appearance
Trevor Canham
Rochester Institute of Technology, Rochester NY; Universitat de Pompeu Fabra, Barcelona
Spain, trevor.canham@upf.edu
Michael J. Murdoch
Rochester Institute of Technology, Rochester NY, michael.murdoch@rit.edu
David Long
Rochester Institute of Technology, Rochester NY, david.long@rit.edu
Written for presentation at the
SMPTE 2018 Annual Technical Conference & Exhibition
Abstract. The share hold of mobile displays in the content distribution market has grown significantly
over the past decade. These displays add new complication to media color management as they can
be viewed across a wide range of environments over a short span of time. There is currently no
consensus within the color science community on the extent to which surround adaptation to ambient
chromaticity has a significant impact on the color appearance of image content on these displays.
Thus, an investigation into this query has been conducted at the Dynamic Visual Adaptation
Laboratory at the Rochester Institute of Technology in Rochester, NY. The study aimed to quantify
the color appearance impact of these surround signals. Observers performed an asymmetric
memory matching task for a set of images viewed under SMPTE standardized mastering conditions
and under a series of ambient illumination conditions with varying chromaticity and luminance. The
results suggested that observers adapt partially to the chromaticity of ambient illumination while
viewing images on portable displays, and also that this mixed adaptation ratio varies as a function of
ambient luminance and stimulus type (self-luminous solid color versus images).
Keywords. Portable Displays, color management, surround adaptation, content mastering
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 2
Introduction
For motion picture distribution, accessibility has proven in many cases to be of greater
importance than picture quality for viewers. In the early 1950s, weekly cinema attendance in the
United States was cut in half as television became popular and accessible to the public. Over
the past decade, history has repeated itself with a decline in television broadcast viewership as
streaming services and portable devices gain in popularity, particularly among younger viewers.
In both examples, consumers chose viewing options which allowed them greater flexibility and
freedom over alternatives that provided better picture quality. Currently, the Society of Motion
Picture and Television Engineers (SMPTE) standards for HDTV mastering are also commonly
used for streaming distribution. In many cases, the assumption is correctly made that viewers
will watch streaming content and television on the same displays and in the same environments.
However, streaming distribution also allows for content to be viewed on mobile platforms that
can be taken into any environment, bringing an additional degree of uncertainty in color
management.
Creating masters for every possible viewing condition in the modern media landscape is
impractical, so any corrective measures would need to be implemented on the device side.
Many manufacturers of these devices have already implemented systems which alter screen
brightness, contrast, and white point in response to changes in surround illumination conditions.
With a well-characterized device, a full-color appearance model could be implemented that
corelates image appearance under varying illumination conditions back to what would have
been seen in the mastering suite.
Far less research has been directed towards chromatic adaptation in this context, in comparison
to other color appearance attributes. The goal of the following research was to model the effect
of chromatic adaptation to surround illumination on the color appearance of image content on
portable displays. The data from this experiment can be useful, if used in conjunction with a
color appearance model, for building an adaptive color management function to be implemented
within these devices. An application of this nature would have practical value for motion picture
professionals and consumers who have an interest in viewing content on portable displays that
exhibits greater fidelity to what might be observed in the mastering suite.
Background
Current standards for HDTV mastering exist in the SMPTE 2080 document series1,2,3, which
sets guidelines for the parameters of reference white luminance level and chromaticity,
standardized viewing environment conditions, and display calibration procedures. These
standards allow for post-production facilities to distribute all mastered HDTV content using a
unified reference color space, allowing for the removal of color management variables when
collaborating between facilities. Most relevant to this research are the display and ambient
luminance values as well as white point chromaticities, which are set to 5 , 100 ,
and D65, respectively. While it is very rare for consumers to set up their viewing environments
and displays in accordance with these standards, they are still helpful for viewers who watch
content regularly on the same displays in the same environment and would prefer a relative
consistency in the mastered look of content from day to day.
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 3
Several studies of the effects of surround adaptation on emissive display color appearance have
been conducted for the application of soft-copy/hard-copy proofing. Gorzynski 4 conducted a
general line of research focused on the topic of achromatic perception in color image displays.
The results of the experiment showed that the adapted state of the observer while performing
an achromatic preference task on a desktop CRT was unaffected by the changing ambient
illumination. Another study was conducted by Henley5 in which corresponding color data in a
soft-copy/hard-copy matching task were analyzed by identifying the von Kries-based6 mixed
adaptation ratio between the display and ambient illumination that best predicted the responses
of observers. The optimized ratios revealed that observers' adapted states were 80%-90%
controlled by the display. A similar study was conducted by Katoh7 in which observers were
found to be roughly 60% adapted to the display. Katoh also chose a von Kries-type mixed
adaptation transform to analyze results. This decision is justified by Bartleson's classification of
chromatic adaptation transforms,8 which shows that von Kries and other similar models are the
most accurate for predicting corresponding colors data for self-luminous stimuli. Likely causes
for the discrepancies between the results of these studies are the experiment tasks, the chosen
stimuli, and the presentation of said stimuli. Thus, in order to gain insight on observer adapted
states while viewing motion picture content on portable displays, an experiment must be
designed which closely reflects the viewing situation in terms of these relevant factors.
Methods
The goal of this experiment is to determine the impact of surround chromatic adaptation on the
color appearance of image content when viewed on mobile displays. As evidenced by the
studies cited above, this requires at the very minimum some number of observers to perform an
achromatic preference or matching task. For the experiment to best quantify the real-life
application of aesthetic color image evaluation on mobile displays, the chosen stimuli will be
images viewed on a device comparable in size to that of a smart phone, tablet, or current-model
on-set reference monitor. The images are first displayed under the SMPTE standard HDTV
mastering illumination conditions as a control. Then, they are displayed again under varying
illumination conditions and observers are asked to work from memory and make adjustments to
match each image back to the control color balance.
This adjustment is executed via a 1-D control scheme which alters a mixed adaptation weighting
factor d in a simple von Kries-based transform, as shown in equation 1. In this equation,
d
represents the user controlled degree of adaptation parameter and
, , , , represent the pseudo-cone fundamentals for the adapted
response, the input, the display white point, and the ambient white point, respectively. These
controls allow for the discretized adjustment of image color balance on straight line vectors in
the 1976 u'v' chromaticity space that intersect the white point of a given ambient condition and
the display white point (Figure 1). With practicality in mind, a single dimension of adjustment
based on this function was determined to be appropriate for this experiment. The majority of
observers would likely not be discerning enough in the reproduction of memorized image color
balance to correct any error in the von Kries predictions and the addition of a second dimension
of control would significantly complicate the task for all observers.
) (1)
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 4
Discrete observer response steps for different ambient vectors are equally spaced in the u’v’
plot by normalizing the input mixed adaptation weighting values by the distance between the
given ambient and display white point. It should be noted that observer responses will be
reported as the chosen mixed adaptation ratio
d
for the remainder of this document and that a
higher
d
value will indicate a greater adaptation to the D65 display white point.
Figure 1. Observer image white point control vectors for each ambient chromaticity setting.
Figure 2. Images used in experimentation, chosen for their neutral and memory color content,
color grading style, and color gamut.
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 5
Ambient illumination chromaticity is not the only variable in this experiment. Image content,
display/ambient brightness ratio, and image starting color balance are also varied between
trials. These additional factors are listed with their chosen levels in table 1. Environmental
illumination is varied between three chromaticity and two brightness levels. To avoid undue
influence of a single image's content on the results, a set of three images is used (Figure 2). A
gray patch serves as a baseline, allowing the results of the experiment to be compared to others
which used solid color stimuli An image of a man and an image of food are used for their
memory color (skin tones, natural food) and significant amount of neutral content. Images are
also presented to observers three times under the same conditions with starting color balances
at different points along the control vector to mitigate presentation bias. Initial color balances for
each ambient chromaticity setting are located on the control vectors in between the display and
ambient points, and on the far ends of the control vectors outside of the display/ambient range.
The experiment was carried out in the Dynamic Visual Adaptation (DVA) Lab at the Munsell
Color Science Laboratory at the Rochester Institute of Technology. The lab is equipped with 14
Philips Color Kinetics LED panels that illuminate the wall behind the display (Figure 4). The
LEDs work on a five-primary system and are capable of producing light of both high spectral
quality and wide dynamic range9. These lights were controlled via 16-bit drive values supplied
by the Matlab experiment test bed, which also displayed images according to experimental
cadence and observer input. Chromatic adaptation transforms were applied to the test images
according to observer input using pre-calculated 3D LUTs. Images generated from the
workstation were displayed via an HDMI connection to a 6.7” (17 cm) Atomos Shogun Flame
portable display. Observers were seated at a viewing distance of 26” (66 cm) from the surface
of the display, resulting in a 7.4 degree vertical viewing angle that reflects a real-world mobile
display viewing scenario. Observers were asked to adjust the height of the seat so that they
were at eye level with the display to avoid any angular display color dependencies.
Table 1. Chosen Experimental Factors
Factor
Type
Levels
Ambient Illumination Color
Varying
Ill A, 2200 K, 10000 K
Ambient Illumination Brightness
Varying
5 nits, 102
Image Content
Varying
Gray patch, man, food
Starting Image Color Balance
Varying
Cool, Middle, Warm
Screen Brightness
Constant
100 nits
Viewing angle
Constant
7.4 Degrees
User task
Constant
Memory Matching
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 6
Table 2. Lighting Settings, averaged over all experiment days
X
Y
Z
U’
V’
D65, 5 nit
4.750
5.004
5.301
0.1985
0.4705
Ill A, 5 nit
5.512
5.054
1.822
0.2541
0.5241
Ill A, 102 nit
112.8
102.1
37.73
0.2566
0.5229
10000k, 5 nit
4.805
5.041
7.359
0.1875
0.4426
10000k, 102 nit
99.69
102.9
151.9
0.1900
0.4412
2200k, 5 nit
6.241
5.041
1.020
0.2939
0.5343
2200k, 102 nit
128.4
104.4
20.18
0.2926
0.5354
Figure 3. Light controlled section of Dynamic Visual Adaptation Lab.
The ambient illumination of the room was characterized using a Photo Research PR 655
Spectrophotometer. Measurements were taken with the device aimed at the wall just above the
display. Normalized tristimulus values and u’v’ chromaticity coordinates were computed within
the device, using the CIE 1931 2Deg standard observer color matching functions. These results
are tabulated above (table 2).
To ensure accurate stimulus presentation and analysis, the display used in the experiment was
characterized in terms of its non-linear tone scale behavior, primary color matrix, dynamic
range, screen flare, and gamut. The display was calibrated to a maximum luminance of 242 nits
so that images could be scaled down to avoid luminance instability as a result of white point
shifts and still meet the 100 nit display standard. The primary matrix was defined using the
chromaticites of the primaries at the scaled maximum code value. To ensure that the image
rendering on the display was consistent across the experiment, a calibration test of the monitor
was performed at the start of each experiment day. A test for out-of-gamut pixels was also
conducted to verify that the monitor was able to reproduce accurate colorimetry of image areas
used for color judgment after adaptation transforms were applied. It was expected that observer
adaptation states were likely to correlate to white points between the monitor and ambient
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 7
chromaticities, so gamut analyses were performed at each of the tested ambient points. It was
found that the essential image evaluation regions (skin tone and achromatic points) remained in
gamut for all tests.
An analysis of observer age, gender, and experience was conducted via a post-experiment
survey. The 24 participants were split 60/40% male/female and had an average age of 28.44
years. When given the opportunity, observers rated themselves at an experience level of
8.52/10 in image evaluation and 7.2/10 in aesthetic color correction. While self-evaluation of this
kind can often be inaccurate, these numbers are reflective of the above-average experience
level of the observer pool as the majority of observers were students and faculty in the film and
color science departments at RIT.
Procedure
Prior to the experiment, all observers were screened for color deficiencies using an Ishihara test
and briefed with experiment instructions. Next, they were led into the experiment area and were
given two minutes to adapt to the SMPTE standard mastering illumination conditions (D65, 5
nit). The test images were then presented in their native color balances and observers were
given as much time as they liked to memorize the appearance of the achromatic and memory
color portions of each image. Next, observers were asked to complete a short training period
introducing them to the experiment task. The initial memorization period was then repeated.
Following this step, the body of the experiment was carried out. The order of illumination
conditions was randomized between observers to avoid any bias in the results caused by the
proximity of the trial to the memorization period. After adapting to each environmental condition,
observers were presented with each of the images from the memorization phase, repeated for
three different initial color balances for a total of nine successive trials. Observers were asked to
adjust the presented image back to their memorized color balance using the control scheme
described in the previous section. Upon completion of all trials, observers were given a post
experiment survey.
Analysis
A test was conducted to measure differences between observer mean mixed adaptation and the
mean for each unique image presentation condition. The average results for each ambient white
point are shown in Figure 4. It can be seen that aside from the case of 10000 K, where
response granularity was low, the average variation between observer responses for the same
permutation of experimental factors was roughly
d = 0.15
. To account for presentation bias in
the experimental results, observers were asked to repeat their color balance decisions from
three different starting points. Figure 5 shows an example histogram of the response ranges
between starting points for each observer and each unique condition when viewed under
Illuminant A. It can be seen that, in many cases, observer responses varied significantly as a
factor of initial image color balance.
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 8
Figure 4. Mean observer mixed adaptation ratio difference from the mean for unique trial
conditions for each environmental white point.
Figure 5. Histogram of mean observer response range for different starting points for the same
stimulus from all Illuminant A ambient trials.
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 9
Figure 6. Mixed adaptation responses for three tested ambient illumination chromaticities
Average observer mixed adaptation ratios between display white point and ambient illumination
chromaticities are plotted with 95% confidence error bars in Figure 6. ANOVA tests were
conducted for the variables for ambient brightness, chromaticity, and stimulus. A significant
difference was found when comparing the mean results from the 10000 K trials to the Illuminant
A and 2200 K trials. This could potentially be a result of the lack of definition in experiment
responses as the distance value between D65 and 10000 K is significantly smaller than that of
the other two ambient white points. Figure 7 shows that the distribution of user responses for
these trials nearly evenly covers the range between
d
values of 0 and 1. A significant difference
was also shown between the user response distributions to the two brightness levels for
Illuminant A and 2200 K. The averages of these distributions are shown in Figure 8. Finally, the
mean mixed adaptation ratio responses for the gray patch stimulus were found to have
statistically significant differences from those of the images (Figure 9).
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 10
Figure 7. Response histograms for all (a) 10,000k, (b) Illuminant A, and (c) 2,200k ambient
illumination trials.
Figure 8. Average responses for different ambient brightness settings.
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 11
Figure 9. Average responses between all trials for given stimuli.
Discussion
Ultimately, the results of this experiment show that some degree of adaptation to the surround
occurs for observers in a portable display viewing application. The adaptation ratio to the display
(roughly
d = 0.7
) for 2200 K and Illuminant A falls in the middle of the results for Katoh7 and
Henly5. To reiterate, this means that observers were approximately 70% adapted to the display
white point under these conditions. The average mixed adaptation response value for the 10000
K setting was found to be significantly lower (roughly 40% adapted to the display). It is possible
that this response variance could be a factor of incomplete adaptation during the warmer
ambient illumination trials.
The effect of brightness on observer adaptation to the surround was also included as a variable
in the experiment conducted by Katoh; however, their results indicated that in doubling the
ambient/display brightness ratio, there was no significant effect on the observer's degree of
adaptation to the surround. On the other hand, the only input data required to compute the
majority of degree of adaptation, mixed, or surround adaptation algorithms is the luminance
level of the adapting stimulus, implying that this is the most relevant factor in this context10.
Based on the results shown in Figure 8, ambient brightness proved to have a statistically-
significant impact on the mixed adaptation ratio of observers under the 2200 K and Illuminant A
ambient chromaticities. This information implies that ambient illumination with a higher
luminance level will cause observers to have a higher degree of adaptation to it. The fact that
the dim ambient luminance level was significantly lower than the display luminance in this
experiment could be a possible explanation for the contrasting results from Katoh's, whose
lowest luminance level was equal to that of the display. These results show that a degree of
adaptation model in this context would need to be both a function of the ambient and display
white chromaticities as well as the ambient/display brightness ratio.
In addition to dependence on the environmental factors of the experiment, a significant variance
was found between the gray patch stimulus and the two image stimuli (Figure 10). This has
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 12
important implications for two reasons. First, the application of mixed adaptation models that are
derived from tests with solid color stimuli cannot be applied in an image stimulus scenario with
reliable predictive accuracy. Second, an adaptive color management application that adjusts a
mixed adaptation ratio based on illumination conditions must also be content-aware. If a system
intends to do all-in-one color management for solid color stimulus (application GUIs, text
applications) as well as image content, white point variation between the two will occur.
Looking at the intra-observer variability results shown in Figure 5, it can be seen that many
observers had a wide range of responses for the same set of presentation factors. In many
cases, this can imply that the observer task was too vague to produce repeatable results.
However, due to the strong signals shown in the general experiment results, as well as the low
mean inter-observer variability shown in Figure 4 for the same presentation conditions, it is
more likely that this high intra-observer variability is a result of presentation bias. This is a
known psychophysical phenomenon, shown also in Fairchild11, in which observers’ responses
seem to gravitate towards the white point of the initial presentation of an image to some degree.
A possible shortcoming of the stimulus presentation in this experiment is the large image areas
which were shifted out of gamut upon applying mixed adaptation transforms. In the post-
experiment survey, participants were asked about the image areas upon which they based their
decisions. A number of observers mentioned using image areas like the man's red hat as well
as the tomatoes and basil from the food image. Since these image areas could not be
proportionally transformed, they represented the adapted appearance of the image less
accurately. This fact has important implications in terms of implementing these types of
transforms in color management. In most cases, transforms of this type will push some amount
of image content out of gamut, which would be counterintuitive to the overall goal of color
appearance preservation. An example using the recommended mixed adaptation ratio from the
2200 K results is shown in Figure 10. It can be seen that areas on the tomatoes and basil are
pushed out of gamut, despite the fact that the image itself is relatively desaturated to begin with.
For a color management application of this nature to legitimately preserve color appearance, it
would need to be implemented on a display platform which includes significant gamut
headroom.
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 13
Figure 10. Gamut test results for food image adapted to 2,200 K with d = 0.7659. Magenta
areas represent out of gamut pixels.
During preliminary trials, some observers noted having the urge to correct the color balance of
the images to their preference over their memory. Observers also noted that the control
presentation of the images did not always align with their color grading preferences, citing
image quality attributes like the saturation of the vegetables and the tone of the man's skin. This
raises an interesting discussion topic as responses in memory matching tasks are, at the very
least, a factor of the observer's ability to memorize and recall the control stimulus color to some
degree. However, the fact that this experiment task involves the color correction of images
brings an additional layer of observer preferential interpretation that may not be present in
single-color memory matching tasks. This must also be factored in when considering the varying
results between the images and the gray patch.
While the large majority of observers reported that the given image adjustment controls were
sufficient, a few observers reported not being able to recreate their memorized image color
balance with the controls provided. This raises the important question of whether or not a
straight line vector in u’v’ space would be a sufficient estimation of an observer's partially-
adapted state. It is entirely possible that some non-linear model might describe observers'
partially-adapted state more accurately than von Kries. A worthwhile continuation of this
research might be to explore the effect of using a novel or alternative chromatic adaptation
model on the experiment results.
Conclusion
This experiment aimed to determine the impact of chromatic adaptation to the surround while
viewing images on portable displays. It was determined that observers had an average mixed
adaption state that was 30%-controlled by the ambient illumination white point. This value was
found to vary as a factor of stimulus type and ambient illumination brightness. Observers
showed a greater adaptation to the surround in cases of both solid color stimuli and brighter
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 14
ambient illumination. Considering the motivating application of this experiment, a worthwhile
expansion of this research would be to test more environmental chromaticity and brightness
levels such that sufficient data could be gathered to build a more complete ‘degree of
adaptation’ model. Additionally, considering the current state of the display and distribution
market, a reformulation of this experiment for HDR/WCG content would make for another
relevant extension. The effects of additional color appearance phenomena like chromatic
induction and luminance adaptation could also be modeled or applied from existing research to
fully relate the appearance of images on mobile displays to their original mastered appearance.
The application of a model of this type in an adaptive device-side solution would be a significant
step in advancing the state of motion picture color management to better preserve creative
intent, regardless of the destination. Hopefully, this document will motivate more research
groups to study these color appearance components and become actively involved with mobile
display color management as it increases in relevance to the media distribution space.
Acknowledgements
Thanks to The School of Film and Animation and The Munsell Color Science Laboratory at the
Rochester Institute of Technology for providing the experiment hardware and location. Thanks
also to Edward Giorgianni, Mark Fairchild, and Marcelo Bertalmio for input on the original
project idea, experiment design, and for additional support. Presentation of this work was
supported by the European Union's Horizon 2020 research and innovation programme under
grant agreement number 761544 (project HDR4EU) and under grant agreement number
780470 (project SAUCE), and by the Spanish government and FEDER Fund, grant ref.
TIN2015-71537-P (MINECO/FEDER,UE).
© 2018 Society of Motion Picture & Television Engineers® (SMPTE®) 15
References
[1] SMPTE ST 2080-1:2014, “Reference White Luminance Level and Chromaticity for HDTV.”
[2] SMPTE RP 2080-2:2014, “Measurement and Calibration Procedures for HDTV Display
Luminance Levels and Chromaticity.”
[3] SMPTE ST 2080-3:2017, “Reference Viewing Environment for Evaluation of HDTV Images.”
[4] Gorzynski, Mark E., "Achromatic perception in color image displays", Thesis, Munsell Color
Science Laboratory, Rochester Institute of Technology, 1992.
[5] Henley, Sharron, "Quantifying mixed adaptation in cross-media color reproduction",Thesis,
Munsell Color Science Laboratory, Rochester Institute of Technology, 2000.
[6] J. von Kries, ‘‘Chromatic Adaptation,’’ Festschrift der Albrecht-Ludwig-Unversitat, Fribourg,
1902 [English translation: D. L. Mac-Adam, Sources of Color Science, MIT Press, Cambridge,
1970].
[7] Katoh, "Effect of ambient light on the color appearance of softcopy images: Mixed chromatic
adaptation for self-luminous displays", Journal of Electronic Imaging 7(4), 794–806, October
1998.
[8] C. J. Bartleson, "Comparison of chromatic-adaptation transforms", Col. Res. Appl. 3, 129-
136, 1978b.
[9] Michael J. Murdoch, "Characterization and control of a multi-primary LED light lab", Opt.
Express 25, 29605-29616, 2017.
[10] CIE, "A Colour Appearance Model for Color Management Systems: CIECAM02" , CIE TC8-
01 Technical Report, CIE Pub. No. 159, 2004a.
[11] Fairchild, "Chromatic Adaptation to Image Displays", Munsell Color Science Laboratory,
Rochester Institute of Technology, 1992
