Content uploaded by Matthias Wölfel
Author content
All content in this area was uploaded by Matthias Wölfel on Oct 07, 2018
Content may be subject to copyright.
Color Preference Differences between
Head Mounted Displays and PC Screens
Andreas Siess
and Matthias W¨
olfel
Faculty of Computer Science and Business Information Systems
Karlsruhe University of Applied Sciences
Karlsruhe, Germany
Email:{andreas.siess, matthias.woelfel}@hs-karlsruhe.de
Nico H¨
affner
Faculty of Digital Media
Furtwangen University of Applied Sciences
Furtwangen, Germany
nico.haeffner@hs-furtwangen.de
Abstract—Recently virtual reality (VR) applications are shift-
ing from professional use cases to more entertainment-centered
approaches. Therefore aesthetic aspects in virtual environments
gain in relevance. This paper examines the influence of different
color determining parameters on user perception habits between
head mounted displays (HMD) and computer screens. We con-
ducted an empirical study with 50 persons that were asked to
adjust the color temperature, saturation and contrast according
to their personal preferences using a HMD as well as a computer
screen, respectively. For cross validation we tested a second
user group of 36 persons that were asked to adjust the color
temperature exclusively. By using a set of five different panorama
images—each of them representing an exemplary scenario—
we have found that color perception differs significantly. This
depends on the used output device as well as gender: i.e. females
preferred a significantly colder color scheme in VR compared to
their preferences on the computer screen. Furthermore they also
chose a significant colder color scheme on the HMD compared to
their male counterparts. Our findings demonstrate that content
created for conventional screens can not simply be transferred
to immersive virtual environments but for optimal results needs
reevaluation of its visual aesthetics.
Index Terms—virtual reality, color preference, cognition, head
mounted display, post processing, perception, color
I. INTRODUCTION
A measurable discrepancy exists between the perception of
real world images and artificially produced pictures—i.e. as
Shim and Lee [1] examined in rendered images that demand
individual post processing and Umbaugh [2] described in
the general work flow with virtual 3D environments. Virtual
Reality (VR) is often seen as a hybrid between imaginary
and realistic media therefore it seems interesting, if and how
the color perception differs from the two mentioned extrema.
This type of content is intuitively assigned to a technological
character—with holistic controlling concepts that are derived
from real world physiological patterns—often described as
natural [3].
Immersive environments therefore demand a new aesthetic
approach that differs in various ways from established me-
dia perception habits. Although the general nature of color
perception in VR can be defined as a synthesis of natural
and artificial interaction and cognition, a simple average mean
estimation between these extrema is not sufficient to determine
the special needs of VR in its complexity. Further examination
could lead to a scientific approach for post processing of con-
tent created via standard creation environments (e.g. computer
screen) for the use in immersive virtual environments.
II. STATUS QUO
One of the very basic introductions to perception of VR
is the chapter Wahrnehmungsaspekte von VR (Aspects of per-
ception in VR) by D¨
orner and Steinicke [4] where they focus
on general phenomena that occur in virtual environments,
although they do not investigate differences between monitor
and head mounted display (HMD) nor observe any color
differences at all. Saleeb [5] however showed that there is
a discrepancy between perception within virtual environments
and the material world in general, albeit he focuses only on
aspects of geometry.
Gusev et al. [6] stated that color and especially color blindness
can have an impact on the prevalence of motion sickness when
using an HMD. Billger et al. [7] examined the requirements
that are essential in order to “make true and realistic color
visualizations in virtual environments”. They found out that it
is necessary to know the color appearance of the real object
when a credible virtual model is desired. They also criticized
the lack of parameters to tweak the color representation of the
virtual objects. Stahre et al. [8] emphasized that there are some
fundamental color reproduction problems that have to be put
into consideration when translating real colors into VR, i.e.
reduced dynamic ranges of screens compared to reality. They
state that this phenomenon is often compensated by virtual
content creators by enhancing contrasts that “go far beyond
what is found in the real world or a realistic reproduction of
it”. Other considerations from software developer Valve [9]
address rendering aspects and optimizations within the device
itself by making use of certain perception phenomena, yet they
also do not put color differences in consideration nor do they
differ between user groups at all.
Outside of VR/HMD there are some studies that show basic
differences of color perception between males and females.
E.g. Hurlbert and Ling [10] demonstrate that males and
females have significantly different preferences for hue values
25
2018 International Conference on Cyberworlds
978-1-5386-7315-7/18/$31.00 ©2018 IEEE
DOI 10.1109/CW.2018.00018
that were not effected by saturation or lightness levels: females
tend to prefer reddish-purple hue values while male rather
prefer blue-greenish hue values. They also state that the
color preferences were more pronounced at the female test
population and lesser visible at the male counterpart. What
also should be put into consideration is the phenomenon of
shifted perception when the color stimulus’ size is changed:
The perceived colorfulness tends to increase when the light
emitting source is enlarged [11].
This rather unsatisfactory summary finds its equivalent in the
quote from McManus at al. which states that the sum of
all color preference studies are “bewildering, confused and
contradictory” [12]—a summary that could be considered
as an indication that although color being an omnipresent
stimulus it lacks a holistic view on its perception.
Fleming [13] showed that color temperatures in artifi-
cial 3d environments are not congruent to their real-world-
counterparts. In order to eliminate this interference source
we decided to test real world images exclusively in order to
determine the perceptual differences between a standardized
PC screen and HMD.
III. TEST SETUP
With these preconditions in mind we designed a test setup
to question the users’ personal preferences—they were asked
to adjust three color-determining parameters until they expe-
rienced a maximum of personal confidence in their individual
preferences in saturation, contrast and color temperature. The
test’s approach was neither the examination of two different
hardware outputs nor to query the technical aspects of its
signal processing: The test’s leading question was whether
there are measurable differences in the individual aesthetic
preference between a PC screen and a HMD.
A. Test Parameters
The following test parameters were examined during the
study:
•color temperature
in steps of 100K from 1.000K (extremely warm color
scheme) to 10.000K (extremely cold color scheme) [14]
•color saturation
in granular steps, correct to four decimal places, values
from 0 to 1, whereas 1 would be the highest possible
saturation [14]
•color contrast
as root mean square (RMS), as shown by Peli [15]
B. Test Scenes
In order to examine the different needs and preferences of
each individual a set of multifaceted images is required to
depict a variety of situations. Obviously it is not possible to
test all thinkable scenarios that could (theoretically) appear
in VR and real life, therefore a categorization of scenarios is
needed. Of each scenario one representative picture is used for
the user tests. Because human vision and the corresponding
cognitive process is also a product of cultural socialization
Fig. 1. The image for the Genre scenario.
[16] the proposed categorization can be found in the different
genres of visual arts. A categorization of different genres,
first introduced by Andr´
eF
´
elibien in 1669—which is still
considered the state of the art in analyzing paintings [17]—is
widely known under the term “Hierarchy of Genres”. F´
elibien
classified the scenarios in four different categories: Landscape,
Portraiture, History and Still Life [18]. This list was later
expanded by the category of Genre-Scenes, probably by Jean-
Baptiste Mass´
e in 1717 [19].
The test scenes (“scenarios”) were captured as 360◦spheri-
cal panoramas as a stitched collage of 26 RAW images, taken
with a Nikon D800E DSLR camera and a Nikkor 14-24mm
f/2.8G ED wide angle lens set to 24mm. During the capturing
process the white balance as well as the exposure time and
dynamic range were measured through the integration of the
Spyder CUBE1on six neuralgic spots around the image. To
ensure the best dynamic range output [21] that the camera is
capable of, all shots were taken at an ISO setting of 50. Since
the situations differed in the lighting conditions we used the
camera’s internal light meter to determine the needed exposure
times. To achieve the best compromise of desired depth of
field (as large as possible) and optical quality (sharpness, low
chromatic aberration, low vignetting) a constant aperture size
of f/8 [22] was selected to take the images. In general we
used rather long (longer than 1/10 sec.) exposure times since
four out of five scenarios were captured at sparsely illuminated
interiors and we used low signal amplification (low ISO) as
well as a comparatively small aperture size.
Figures 1 to 5 show the used scenarios as two dimensional
images in their basic, unaltered version (full dynamic range,
neutral white balance).
C. Test approach and setup
For designing the test application we used the game engine
Unity with the basic, unaltered SteamVR Plugin [23] as
framework. As Figure 6 shows: the captured PNGs were used
as inner texture on a 3D sphere where each individual test
1The Spyder CUBE can be described as an extended gray card that can be
used to “set the white balance, exposure, black level and brightness” [20] in
an image.
26
Fig. 2. The image for the History scenario.
Fig. 3. The image for the Landscape scenario.
Fig. 4. The image for the Portrait scenario.
Fig. 5. The image for the Still Life scenario.
Fig. 6. Scheme of the test scene in Unity: The spectator stands inside a 3D
sphere where the tested image is displayed as texture of the sphere’s inner
surface. The spectator can rotate freely, yet he is not able to move.
person was placed inside as spectator2. In this environment
the probands had 360° freedom in every rotation angle while
their position within the test scene remained fixed. In the test
row where the HMD was used as output device, the virtual
spectator’s rotation was attached to the headset’s physical
rotation. In order to enable a comparable interaction when
testing the PC/standard screen setup, the virtual camera could
be rotated with a mouse. In the HMD setup the test person
could manipulate the parameters by using a Xbox One wireless
controller whereas the PC screen setup made use of a stan-
dard PC keyboard. The control of the application was kept
comparatively simple, therefore it was sufficient to explain
the operation briefly to the participants before the testing
procedure, especially since the test supervisor was in the room
permanently and available for questions regarding the controls.
Since this test configuration was designed to simulate standard
content creator’s working conditions we used three norms as
guideline how to build up the workplace layout: More con-
cretely we used DIN 15996 [26] to get the normalized distance
from spectator to screen and the tilt angle of the screen itself.
From DIN 5034-1 [27] we extracted the information about the
recommended ambient brightness in the testing environment
and the minimum size of the windows as well as the position
of the working desk within the room. Finally we used DIN
EN 527-1 [28] for the height of the used desk and chair.
By using this setup we tried to simulate realistic content
creation conditions that can produce meaningful results for the
professional practice. By specifically scheduling our testing
periods we ensured that the overall spatial color temperature
at the working desk was in the range of 4500 K to 6700 K
while the ambient brightness met the requirements of DIN
5034 [27].
The users were asked to adjust the parameters all together
and could correct their decisions until they were satisfied
2For the whole test procedure we used Unity’s linear color space [24]
a subtype of sRGB that is considered as standard display color space that
is—with some limitations—roughly device independent [25].
27
with their result3. In that case they told the instructor to save
the chosen parameters and the next scenario was loaded. In
order to avoid the consequences of priming [29] the following
parameters were randomized: If the test subject had to start
with PC screen or HMD; the sequential order of the scenarios
and each of the starting parameters of each scenario. For the
whole testing period we used the following gear as input and
output devices:
•Dell UP2516D 25” TFT. Input method: keyboard and
mouse
•HTC Vive. Input method: Xbox One wireless controller
The HMD-setup followed the basic recommendations that the
Vive’s user manual provided for room-scale-experiences [30].
Although the users were not able to move inside the immersive
virtual environment—except looking around—we granted a
3×3m wide area where the test persons could operate the Vive
safely.
D. Screen Calibration
In order to ensure a comparable color output between the
two used devices a calibration process is needed to ensure
a consistent color output throughout the tests. The computer
screen was calibrated by using a DSLR Camera (Nikon
D800E) and a fixed focus macro lens (Nikkor 60mm f/2.8).
The brightness of each screen was adjusted by measuring
the lightness output through the camera’s internal light meter
(selected mode was matrix metering). All adjustments were
made by tweaking the computer screen settings, since it is
not easily possible to access the Vive’s image settings on a
hardware-close level.
In order to calibrate the color output between the two screens
amiddle gray [31] [25] was used that was displayed over
the whole screen while the camera was adjusted to picture
the whole frame. The screen settings were adjusted through
a successive process until both captured images were abso-
lutely identical (this was measured through Photoshop’s divide
function). In addition to this procedure we also used the exact
images (Figures 1 to 5) that would be utilized afterwards in
the testing process.In order to shield the lens from ambient
light—which would result in errors through flare—a 100%
light-blocking fabric was used that covered the whole setup
(camera plus screen/Vive). Moreover the test room was only
dimly lit during the calibration process and the test procedure
was done during the night. Since the test took place over the
course of several days we established a review process of the
found screen presets to ensure a consistent output. However
it turned out that neither the PC screen nor the Vive differed
significantly from the first found parameters.
E. Test Population
During the first test period 50 subjects were tested, of which
28 persons were male and 22 persons were female. The age
3For this process the test persons could take as much time as they wanted—
we decided to avoid giving any time guidelines in order to guarantee that the
participants could alter the parameters without pressure until their personal
preference was found.
TABLE I
AVERAGE COLOR SATURATION PREFERENCES OF THE DIFFERENT
SCENARIOS.THE STANDARD DEVIATION IS LISTED IN BRACKETS.
female male
screen HMD screen HMD
Genre 0.38 (0.14) 0.39 (0.21) 0.38 (0.14) 0.44 (0.25)
History 0.36 (0.21) 0.26 (0.18) 0.25 (0.15) 0.37 (0.22)
Landscape 0.20 (0.14) 0.36 (0.18) 0.19 (0.13) 0.21 (0.13)
Portrait 0.36 (0.13) 0.46 (0.20) 0.36 (0.16) 0.42 (0.26)
Still Life 0.29 (0.14) 0.29 (0.20) 0.35 (0.29) 0.37 (0.25)
TABLE II
EFFECT SIZES OF DIFFERENCES IN COLOR SATURATION BETWEEN
DIFFERENT OUTPUT DEVICES,CALCULATED IN COHEN’SD.AN ASTERISK
INDICATES A MEDIUM,TWO ASTERISKS A LARGE EFFECT SIZE.
female male
Genre 0.74** 0.29
History 0.49* 0.68*
Landscape 1.05** 0.20
Portrait 0.57* 0.28
Still Life <0.01 0.07
range differed from 18 to 31 years with a mean age of 24.
The second test stage featured 36 subjects of which 16 persons
were male and 20 persons were female. The age range differed
from 19 to 61 years with a mean age of 36. Across all tests,
67.6% of women had previous experience in VR, whereas only
59.7% of the male participants had already used a HMD.
IV. RESULTS
In general it can be stated that there is a measurable
difference between preferred color settings for HMD and PC
screen respectively as well as across the genders.
In the following paragraphs we calculated the effect sizes with
the pooled standard deviation as Cohen’s d [32] using this
formula:
d=M2
1−M2
2
(n1−1)s2
1+(n2−1)s2
2
n1+n2−2
n = sample size; s = standard deviation; M = sample mean
A. Saturation
All test persons preferred a more saturated image while
using a HMD compared to the PC screen. As Table I shows
the biggest differences occurred at the scenario Landscape
with the female test participants. The closest distance can be
observed within the female population at the Still Life scenario.
As shown in Table II the Landscape scenario features also the
largest effect size. The results in color saturation also featured
the only time where the male effect size was larger than the
female counterpart (scenario History).
B. Contrast
As seen in Table III it can be stated that females preferred a
higher-contrast image in the HMD compared to the PC screen.
This effect is also visible within the male test population, with
two exceptions being the Landscape and Portrait Scenario. As
shown in Table IV the effect sizes are significantly lower than
the values for color temperature.
28
TABLE III
AVERAGE CONTRAST (RMS) PREFERENCES OF THE DIFFERENT
SCENARIOS.THE STANDARD DEVIATION IS LISTED IN BRACKETS.
female male
screen HMD screen HMD
Genre 0.13 (0.03) 0.15 (0.02) 0.15 (0.03) 0.15 (0.03)
History 0.12 (0.04) 0.14 (0.05) 0.12 (0.03) 0.12 (0.04)
Landscape 0.10 (0.03) 0.11 (0.03) 0.10 (0.02) 0.10 (0.02)
Portrait 0.13 (0.04) 0.14 (0.04) 0.14 (0.03) 0.14 (0.04)
Still Life 0.21 (0.07) 0.25 (0.08) 0.20 (0.04) 0.22 (0.07)
TABLE IV
EFFECT SIZES OF DIFFERENCES IN COLOR CONTRAST BETWEEN
DIFFERENT OUTPUT DEVICES,CALCULATED IN COHEN’SD.AN ASTERISK
INDICATES A MEDIUM,TWO ASTERISKS A LARGE EFFECT SIZE.
female male
Genre 0.71** 0.22
History 0.30 0.23
Landscape 0.33 0.16
Portrait 0.32 0.12
Still Life 0.53* 0.35
C. Color Temperature
One of the most relevant learnings was the preference for
cooler color temperatures of the females in VR applications
(HMD, mean color temperature: 8.003K) compared to the PC
screen (mean: 5.988K). As Table V shows the differences
at the female test group were quite significant, with the
Still Life scenario being the only exception, although this
phenomenon did not occur within the male test population, on
contrary: They even preferred slightly warmer color schemes
on some scenarios but were in general very close to their
preferred parameters at the PC screen. It can also be stated that
even the distance within the females was considerably large:
While using the PC screen they chose mostly an significantly
warmer color scheme which leads to an even bigger divergence
between these two output devices.
Also males prefer slightly warmer color settings in the
HMD (mean color temperature: 6.600K) compared to the PC
screen (mean color temperature: 6.794K). These diametrically
opposed color preferences between males and females is also a
strong indication that the technical color calibration described
in chapter III-D was unbiased. As shown in Table VI it can
be stated that the effect sizes of the females are either large
(scenario Portrait or Genre) or—interpreted by the recommen-
dations of Sawilowsky [33]—very large (scenario Landscape
TABLE V
AVERAGE COLOR TEMPERATURE PREFERENCES (IN KELVIN)OF THE
DIFFERENT SCENARIOS.THE STANDARD DEVIATION IS LISTED IN
BRACKETS.
female male
screen HMD screen HMD
Genre 6318 (2400) 8563 (2716) 6358 (1683) 6806 (2345)
History 5709 (1675) 8127 (2211) 6717 (1873) 5800 (1740)
Landscape 6136 (2219) 8963 (2143) 7547 (2340) 7386 (2647)
Portrait 5645 (2321) 8109 (2749) 7586 (2371) 7296 (2594)
Still Life 6136 (1982) 6254 (1720) 5764 (2041) 5714 (1910)
TABLE VI
EFFECT SIZES OF DIFFERENCES IN COLOR TEMPERATURE BETWEEN
DIFFERENT OUTPUT DEVICES,CALCULATED IN COHEN’SD.AN ASTERISK
INDICATES A MEDIUM,TWO ASTERISKS A LARGE AND THREE ASTERISKS
A VERY LARGE EFFECT SIZE.
female male
Genre 0.90** 0.22
History 1.26*** 0.52*
Landscape 1.33*** 0.07
Portrait 0.99** 0.13
Still Life 0.07 0.03
and History). As mentioned before the only exception is the
scenario Still Life. The male effect size is small (scenario
Genre)tomedium (scenario History) at best, whereas the other
scenarios featured barely any measurable differences.
D. Cross-Validation through a second user study
Since we found a significant difference for the color
temperature only and to ensure this effect is not due to
the variation of the other parameters we decided for a
post-hoc test: we altered our test setup and asked additional
36 test persons to adjust the color temperature exclusively
while the contrast as well as the saturation remained in a
neutral position. With this alteration we expected a validation
of our first test setup if the measured color temperature
discrepancies were a valid standalone phenomenon and not an
artifact/cross-reaction caused by the other tested parameters.
We also extended the test group population4to a more senior
user group with a mean age of 37 (females) respectively 36
(males). As shown in Table VII our results from the first test
were not only confirmed but the differences are even better
pronounced. On the screen females tend to prefer a significant
warmer environment (mean over all scenarios: 5005 K)
compared to their preferences for the HMD (mean over all
scenarios: 9333 K). Regarding the male test population we
also found that the tendency towards warmer environments
in VR compared to PC screens could be validated. Males
preferred a significant warmer environment for the HMD
(mean over all scenarios: 4876 K) compared to PC screens
(mean over all scenarios: 6654 K). Even in scenarios where
the results were quite ambiguous in the first test setup (e.g.
Genre or Still Life) we measured quite conclusive values in
the second testing setup.
This second test showed that color temperature preferences
differ from gender as well as from the used output device,
regardless of the parameters contrast and saturation. The effect
size of this phenomenon was even larger when the test persons
were asked to adjust this parameter in an isolated way. Table
VIII compares the color temperatures’ effect sizes between
the two test setups: At the second test row all effect sizes
were considerably large with two exceptions at the male test
population: the scenarios Genre and Portrait featured only
4We tested completely different persons—none of the test population of the
first test did take part in the second experiment.
29
TABLE VII
AVERAGE COLOR TEMPERATURE PREFERENCES OF THE DIFFERENT
SCENARIOS WHEN ASKED TO ALTER THE PARAMETER COLOR
TEMPERATURE IN ISOLATED MANNER (2ND USER STUDY). THE STANDARD
DEVIATION IS LISTED IN BRACKETS.
female male
screen HMD screen HMD
Genre 4187 (1290) 8773 (3068) 5872 (2356) 4600 (1943)
History 4880 (1622) 9373 (2940) 6309 (2458) 4327 (1536)
Landscape 5427 (1681) 9920 (2590) 7854 (2759) 5181 (1879)
Portrait 5920 (2527) 10400 (1654) 6545 (2630) 5436 (2354)
Still Life 4613 (1357) 8200 (3311) 6691 (1950) 4836 (1355)
TABLE VIII
COMPARISON OF EFFECT SIZES,CALCULATED AS COHEN’SD,ACROSS
THE TEST ROWS FOR THE PARAMETER color temperature BETWEEN HMD
AND PC SCREEN.TEST ROW 1DEMANDED AN ADJUSTMENT OF ALL
THREE PARAMETERS,WHILE TEST ROW 2ONLY QUESTIONED THE COLOR
TEMPERATURE PREFERENCE EXCLUSIVELY
.AN ASTERISK INDICATES A
MEDIUM,TWO ASTERISKS A LARGE,AND THREE ASTERISKS A VERY
LARGE EFFECT SIZE.
female male
1st test 2nd test 1st test 2nd test
Genre 0.90** 1.95*** 0.22 0.59*
History 1.26*** 1.89*** 0.52* 0.97**
Landscape 1.33*** 2.06*** 0.07 1.13***
Portrait 0.99** 2.09*** 0.13 0.44
Still Life 0.07 1.42*** 0.03 1.12***
medium effect sizes. In general the clarity of the desired
preferences were quite remarkable. As a second comparison
we decided to do an analysis of variance (ANOVA) [34] to
compare the results of test 1 with test 2 of the output device’s
impact on the color temperature preference. As Table IX
shows: Also the ANOVA tests confirm the large effect size of
the output device on the personal color temperature preference,
especially at the female part of the test population.
TABLE IX
VALUES OF P AS RESULTS OF AN ANOVA ON THE INPUT DEVICE’S
IMPACT ON THE COLOR PREFERENCE.
female male
1st test 2nd test 1st test 2nd test
Genre 0.007 <0.001 0.407 0.182
History <0.001 <0.001 0.059 0.035
Landscape <0.001 <0.001 0.806 0.015
Portrait 0.003 <0.001 0.628 0.310
Still Life 0.834 <0.001 0.925 0.018
E. Visual Output
Since in the second test row the participants were asked to
adjust the parameter color temperature exclusively a detailed
consideration should be put on how the output actually looked
like for the users. The collages show a snippet of each scenario
respectively—as an introduction a non-altered version of the
captured image (as it was used as basis) is also included.
Fig. 7. Results from the 2nd test row: Visual output of the Genre scenario
(snippets).
Fig. 8. Results from the 2nd test row: Visual output of the History scenario
(snippets).
Figure 7 shows that in comparison to the standard image5
both genders preferred a slightly warmer image on the com-
puter screen whereas females altered the image significantly
into a colder appearance while using a HMD. However this
effect did not occur at the male test population. Figure 8
shows a comparable phenomenon, however the preferred color
scheme of the males for the HMD is a lot warmer than its
monitor counterpart. Although Figure 9 shows comparable
values it is obvious that all test persons chose warm color
schemes at the PC screen (especially visible at the sky tones),
whereas females altered the image to a significantly colder
color output on the HMD.
5Although it has to be emphasized that none of the test person ever saw this
unaltered image. When they began the testing procedure they were presented
with randomized settings of all three parameters to force them into a cognitive
process to recapitulate their truly preference and not only to accept the given
image due to compromise.
Fig. 9. Results from the 2nd test row: Visual output of the Landscape scenario
(snippets).
30
Fig. 10. Results from the 2nd test row: Visual output of the Portrait scenario
(snippets).
Fig. 11. Results from the 2nd test row: Visual output of the Sill Life scenario
(snippets).
Figure 10 also shows the clear tendency of the female
participants to shift the image in the HMD to the colder
color range, whereas the male test subjects preferred a warmer
image. In general, it can also be stated that there were clearly
visible deviations from the original image (on the PC monitor
as well as in the HMD). None of the test groups was satisfied
with an unchanged picture. While Still Life could be seen as a
big exception of the found phenomena in the first test row it
can now be stated that the effect of shifted color temperature
values also appears in this scenario when the second test row
is put into consideration. Figure 11 shows: Female test persons
tend to prefer a colder color scheme on the HMD than on the
computer screens whereas males shifted the parameters into
the warmer region when compared to their preferences for the
PC screen.
V. D ISCUSSION AND FURTHER RESEARCH
Human color perception is strongly influenced by the
variance of surround colors: As Brown and MacLeod [35]
stated that objects appear much more vivid when observed in
a low-contrast, gray surrounding compared to a high-contrast,
colorful environment. Since the Vive does not yet offer a field
of view (FOV) that is congruent with human vision (FOV
Vive: 110° [36] compared to human FOV (horizontal) of
200° [37]) the conditions inside an HMD seem comparable
to a surrounding with faded, gray colors and low contrast
that Brown/MacLeod described, since the inner edge of the
headset is always within the user’s peripheral field of view.
In the first test row we found that users preferred a more
saturated image in VR compared to the neutral baselines
(Figures 1 to 5). On the one hand this overcompensation
could be explained by the user’s expectation to oversaturated
colors within low contrast surroundings. On the other hand
this phenomenon could also be seen as an indicator that
the findings of Stahre et al. [8] are still correct: Users
expect vivid, colorful applications in VR because they are
used to the trend of content creators to compensate the
unsatisfactory dynamic range of HMDs by oversaturating
the displayed images. Outside of VR there are some studies
that show a difference between female and male color
preferences: i.e. Hurlbert and Ling [10] showed that European
females’ preferences were shifted significantly towards the
“reddish-purple region” while males showed a blue-green
bias. Although our study did not question isolated hue values
we can still state that we could observe an opposing effect
in our results when using a HMD: Females preferred color
schemes that are fundamentally different to the findings from
Hurlbert/Ling—although this study shows comparable gender
differences: the results of the male test group were also lesser
pronounced than the female part. Finally, what also should
be put into consideration are perceptual phenomena like the
Gelb Effect [38]: The perceived lightness of an object is
significantly stronger when the observed item is surrounded
by a darker environment. As Stewart [39] stated: This effect
should be revealed by differences in contrast—what also
could not be observed in our study. We also consider to
alter our test setup towards 3D environments where the test
persons are able to move through virtual environments in
order to ensure a second cross validation of our results.
VI. POSSIBLE CONSEQUENCES:HOW TO ADDRESS THESE
PHENOMENA IN PRODUCTION
Most of the content for HMDs is still created via standard
computer screens [40]: Our findings suggest that a color
correction is necessary once the application is transferred
from the development environment to a HMD. In general
this post-processing could either already be addressed
while developing the application or lastly at the end user’s
preferences setup. Both approaches have its advantages
and disadvantages: i.e. it is possible to take account of the
differing color scheme at early stage when including the
color correction into the development process, though the
application’s target audience must be known and clear-cut.
Depending on the chosen approach different methods are
thinkable to realize this proposed color correction: On a
hardware-close level screen calibration presets could be used
that automatically shift all perceived colors to neutral values
during development. Although this solution could be easily
implemented we propose a different approach that allows
A/B comparisons on the fly and can be used on developer’s
as well as on user’s side: Look-up-tables (LUT) have proven
that they are capable to provide potent color calibration
possibilities [41] and can be used inside of popular game
engines as a camera effect [42] [43] or with other content
creation tools (i.e. Photoshop) as image effect.
31
VII. CONCLUSIONS
While various research has demonstrated that there are
differences in perception between immersive VR environments
and reality (e.g. dimension [5]) as well as screens (e.g.
cyber sickness [44]) immersive environments therefore require
not only rethinking storytelling and interaction, but also re-
considering aesthetics as an independent factor that breaks
with established media perception habits. These individual
preferences especially come apparent when examining the
color temperature of an image. We have shown that females
prefer significantly colder color schemes in general whereas
males tend to favor an almost identical image using a HMD
compared to PC screens.
REFERENCES
[1] H. Shim and S. Lee, “Automatic color realism enhancement for virtual
reality,” pp. 59–60, March 2012.
[2] S. E. Umbaugh, Computer vision and image processing:apractical
approach using CVIPtools. Upper Saddle River, NJ : Prentice Hall
PTR, 1998.
[3] M. R. Mine, F. P. Brooks, Jr., and C. H. Sequin, “Moving objects in
space: Exploiting proprioception in virtual-environment interaction,” in
Proceedings of the 24th Annual Conference on Computer Graphics and
Interactive Techniques, ser. SIGGRAPH ’97. New York, NY, USA:
ACM Press/Addison-Wesley Publishing Co., 1997, pp. 19–26.
[4] R. D¨
orner, W. Broll, P. Grimm, and B. Jung, Virtual und Augmented
Reality (VR / AR): Grundlagen und Methoden der Virtuellen und
Augmentierten Realit¨
at, ser. eXamen.press. Springer Berlin Heidelberg,
2014.
[5] N. Saleeb, “Effects of the differences between virtual and physical
perception of space on building information modelling,” pp. 21–32,
2015.
[6] D. A. Gusev, D. M. Whittinghill, and J. Yong, A Simulator to Study
the Effects of Color and Color Blindness on Motion Sickness in Virtual
Reality Using Head-Mounted Displays. Singapore: Springer Singapore,
2016, pp. 197–204.
[7] M. Billger, I. Heldal, B. Stahre, and K. Renstrom, “Perception of color
and space in virtual reality: a comparison between a real room and virtual
reality models,” in Electronic Imaging 2004. International Society for
Optics and Photonics, 2004, pp. 90–98.
[8] B. Stahre and M. Billger, “Physical measurements vs visual perception:
Comparing colour appearance in reality to virtual reality,” in Conference
on Colour in Graphics, Imaging, and Vision, vol. 2006, no. 1. Society
for Imaging Science and Technology, 2006, pp. 146–151.
[9] A. Vlachos, “Advanced vr rendering,” Valve, March 2015, accessed
04-20-2018. [Online]. Available: http://alex.vlachos.com/graphics/Alex
Vlachos Advanced VR Rendering GDC2015.pdf
[10] A. C. Hurlbert and Y. Ling, “Biological components of sex differences
in color preference,” Current Biology, vol. 17, no. 16, 2007.
[11] P. Bodrogi and T. Khan, Illumination, color and imaging: evaluation
and optimization of visual displays. John Wiley & Sons, 2012.
[12] I. C. McManus, A. Jones, and J. Cottrell, “The aesthetics of colour,”
Perception, vol. 10, no. 6, pp. 651–666, 1981.
[13] B. Fleming, 3D photorealism toolkit. John Wiley & Sons, Inc., 1998.
[14] E. L¨
ubbe, Farbempfindung, Farbbeschreibung und Farbmessung: Eine
Formel f¨
ur die Farbs¨
attigung, ser. SpringerLink : B¨
ucher. Springer
Fachmedien Wiesbaden, 2012.
[15] E. Peli, “Contrast in complex images,” JOSA A, vol. 7, no. 10, pp.
2032–2040, 1990.
[16] A. Braun, Wahrnehmung von Wald und Natur, ser. Forschung Soziolo-
gie. VS Verlag f ¨
ur Sozialwissenschaften, 2013.
[17] C. Bailey, P. Conisbee, T. Gaehtgens, N. G. of Canada, N. G.
of Art (U.S.), and G. Gem¨
aldegalerie (Berlin, The Age of Watteau,
Chardin, and Fragonard: Masterpieces of French Genre Painting. Yale
University Press, 2003.
[18] A. Felibien, Conferences de l’Academie Royale de peinture et de
sculpture. Estienne Roger, 1706.
[19] T. Kirchner, “Paradigma der Gegenw¨
artigkeit. Schlachtenmalerei als
Gattung ohne Genrekonventionen,” Bilder der Macht, Macht der
Bilder—Zeitgeschichte in Darstellungen des 19. Jahrhunderts, pp. 107–
127, 1997.
[20] datacolor, “Spydercube—control color by balancing light,” 2017,
accessed 10-20-2017. [Online]. Available: http://www.datacolor.com/
photography-design/product- overview/spydercube/
[21] DxO, “Nikon D800: The best sensor analyzed on dxomark!” 2012,
accessed 10-22-2017. [Online]. Available: https://www.dxomark.com/
Reviews/Nikon-D800-Review/Nikon-D800- Sensor-performance
[22] ——, “Nikon AF-S NIKKOR 14-24mm f/2.8G ED mounted
on Nikon D800: Measurements,” 2013, accessed 10-22-2017.
[Online]. Available: https://www.dxomark.com/Lenses/Nikon/Nikon-
AF-S- NIKKOR-14-24mm-F28G- ED-mounted- on-Nikon- D800
[23] Valve Corp., “Steamvr plugin, version 1.2.2,” accessed 12-09-2017.
[Online]. Available: https://www.assetstore.unity3d.com/en/#!/content/
32647
[24] Unity, “Documentation 2017: Linear rendering overview,” accessed
12-09-2017. [Online]. Available: https://docs.unity3d.com/Manual/
LinearLighting.html
[25] P. Bodrogi and T. Q. Khanh, “Colorimetric and color appearance-
based characterization of displays,” Illumination, Color and Imaging:
Evaluation and Optimization of Visual Displays, pp. 25–95, 2012.
[26] DIN-15996, “Image and sound production in film and video studios and
radio stations – principles and provisions for a work station,” May 2008.
[27] DIN-5034-1, “Daylight in interiors – part 1: General requirements,” July
2011.
[28] DIN-EN527-1, “Office furniture – work tables and desks – part 1:
Dimensions,” August 2011.
[29] P. Herr, S. Sherman, and F. Russell, “On the consequences of priming:
Assimilation and contrast effects,” Journal of Experimental Social
Psychology, vol. 19, no. 4, pp. 323 – 340, 1983.
[30] Valve Corp., “Htc vive installation guide,” 2016, accessed 10-22-2017.
[Online]. Available: https://support.steampowered.com/steamvr/HTC
Vive /
[31] E. Reinhard, W. Heidrich, P. Debevec, S. Pattanaik, G. Ward, and
K. Myszkowski, High Dynamic Range Imaging: Acquisition, Display,
and Image-Based Lighting, ser. Morgan Kaufmann series in computer
graphics. Elsevier Science, 2010.
[32] J. Cohen, Statistical Power Analysis for the Behavioral Sciences. Taylor
& Francis, 2013.
[33] S. Sawilowsky, “New effect size rules of thumb.” Journal of Modern
Applied Statistical Methods, vol. 8, no. 2, p. 467–474, 2009.
[34] J. Lazar, J. H. Feng, and H. Hochheiser, Research methods in human-
computer interaction. Morgan Kaufmann, 2017.
[35] R. O. Brown and D. I. MacLeod, “Color appearance depends on the
variance of surround colors,” Current Biology, vol. 7, no. 11, pp. 844–
849, 1997.
[36] Valve Corp., “Vive specs: Headset specs, 2017,” accessed 12-
06-2017. [Online]. Available: https://www.vive.com/us/product/vive-
virtual-reality- system/
[37] E. B. Werner, Manual of visual fields. Churchill Livingstone, 1991.
[38] A. Gelb, “Die Farbenkonstanz der Sehdinge,” in Handbuch der normalen
und pathologischen Physiologie. Springer, 1929, pp. 594–678.
[39] E. C. Stewart, “The gelb effect.” Journal of Experimental Psychology,
vol. 57, no. 4, p. 235, 1959.
[40] J. Carmack, “The dawn of mobile vr. gdc 2015, san francisco, ca
(2015),” accessed 11-29-2017. [Online]. Available: http://www.gdcvault.
com/play/1022304/The-Dawn-of-Mobile
[41] P.-C. Hung, “Colorimetric calibration in electronic imaging devices
using a look-up-table model and interpolations,” Journal of Electronic
imaging, vol. 2, no. 1, pp. 53–62, 1993.
[42] Unity, “Documentation 2017: User lut,” accessed 12-06-2017.
[Online]. Available: https://docs.unity3d.com/Manual/PostProcessing-
UserLut.html
[43] Unreal, “Engine documentation: Color grading,” accessed 12-06-2017.
[Online]. Available: https://docs.unrealengine.com/latest/INT/Engine/
Rendering/PostProcessEffects/ColorGrading/
[44] A. Sieß, S. Beuck, and M. W¨
olfel, “Virtual reality–quo vadis? How
to address the complete audience of an emerging technology,” in
Collaborative European Research Conference 2017, 2017, pp. 245–247.
32
