Conference PaperPDF Available

Physiologically Personalized Color Management for Motion Picture Workflows

Authors:

Abstract

One of the essential mechanisms employed by the human visual system when interpreting the natural world is that of trichromatic integration of physical scene spectra by cone photoreceptors. By extension of this, different scene spectra can result in the same color sensation in an observer, a phenomenon known as metamerism. This allows imaging systems to produce realistic reproductions of scene content by the same three channel mechanism. To predict these matches, color matching functions (CMFs) are used which aim to describe the average spectral integration behavior of observers. However, the use of a single average observer CMF has been shown to result in impactful color rendering errors, as there exists significant variation in the spectral absorption characteristics of the eye within populations of color-normal observers. When this is crossed with the growing disparity between the spectral characteristics of emerging display technology it becomes evident that this inter-observer variability should be accounted for. Asano and Fairchild present a physiologically based individual observer model, as well as a method for separating a population of observers into a limited number of categorical CMFs. Building on this work, we present a computationally simple metameric match simulation pipeline which uses these categorical functions. With this pipeline, we perform a simulation with real display spectra and natural images to observe the variability which could occur among a population as a result of observer metamerism in a motion picture viewing scenario. The results provide further evidence that inter-observer metameric variability is a relevant problem in the context of natural images. Finally, we outline how this pipeline can be incorporated into one's color management strategy.
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. 2020. 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).
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®)
SMPTE ATC 2020
Physiologically Personalized Color Management for
Motion Picture Workflows
Trevor D. Canham
Universitat Pompeu Fabra, Barcelona, Spain
David L. Long
Rochester Institute of Technology, Rochester NY, USA
Mark D. Fairchild
Rochester Institute of Technology, Rochester NY, USA
Marcelo Bertalmío
Universitat Pompeu Fabra, Barcelona, Spain
Written for presentation at the
SMPTE 2020 Annual Technical Conference & Exhibition
Abstract. One of the essential mechanisms employed by the human visual system when interpreting
the natural world is that of trichromatic integration of physical scene spectra by cone photoreceptors.
By extension of this, different scene spectra can result in the same color sensation in an observer, a
phenomenon known as metamerism. This allows imaging systems to produce realistic reproductions
of scene content by the same three channel mechanism. To predict these matches, color matching
functions (CMFs) are used which aim to describe the average spectral integration behavior of
observers. However, the use of a single average observer CMF has been shown to result in impactful
color rendering errors, as there exists significant variation in the spectral absorption characteristics of
the eye within populations of color-normal observers. When this is crossed with the growing disparity
between the spectral characteristics of emerging display technology it becomes evident that this inter-
observer variability should be accounted for.
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. 2020. 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).
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®)
Asano and Fairchild present a physiologically based individual observer model, as well as a method
for separating a population of observers into a limited number of categorical CMFs. Building on this
work, we present a computationally simple metameric match simulation pipeline which uses these
categorical functions. With this pipeline, we perform a simulation with real display spectra and natural
images to observe the variability which could occur among a population as a result of observer
metamerism in a motion picture viewing scenario. The results provide further evidence that inter-
observer metameric variability is a relevant problem in the context of natural images. Finally, we outline
how this pipeline can be incorporated into one’s color management strategy.
Keywords. Observer metamerism, color management, personalization.
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 2
Introduction
Through evolution, the discrimination abilities of our visual system have gradually developed
according to our survival needs. This process started with simple bright versus dark discrimination
in our single celled ancestors, adding spatial discrimination, color, etc. to eventually create the
visual sensation that we rely on constantly, complete with its high level information management
routines and constantly dynamic behavior. By virtue of its construction, this visual experience is
an abbreviated probe of physical reality. For example, instead of discriminating one color from
another based on their specific physical spectra, we compare the relative integrated responses
of three retinal cell types (cones) which are sensitive to different bands of the spectrum. As a
result of this mechanism, these three cones could produce the same sensation from stimuli that
are spectrally different, a concept we refer to as metamerism.
It is because of metamerism that imaging systems can be engineered which are able to convey
a certain degree of realism similarly using just three color channels. This allows camera and
display engineers to employ solutions that involve the modulation of the spectral characteristics
of primaries to achieve various performance goals (dynamic range and color gamut, color
rendering consistency, energy efficiency, etc.) In this sense, the precise spectral shape is
percieved to be a visually non-impactful parameter, so long as it results in a matching cone
response for the observer (compared to some reference representation). The problem in this
strategy is that while the reference observer may percieve a match, variation in the physiology of
the human eye within a population makes it such that not everyone will recieve the same
sensation as a result of differing spectral interaction between scene/screen light and the eye.
Furthermore, if the description of the reference observer does not match that of any real observer
(ie. if it is a population average), there exists the possiblity that no viewer would see the correct
color rendering.
In lieu of more spectrally accurate imaging systems, a color management pipeline for motion
picture production which considers the individual characteristics of key creative observers could
remove observer metamerism variability when color critical decisions are being made. For
example, if for the various stages at which a Director of Photography reviews imagery (on-set,
dailies, color correction, VFX, etc.) their monitoring equipment is calibrated to be a metameric
match to some reference considering their individual physiology, this would remove inconsistency
relating to the display technology used at different stages and avoid any influence it may have on
creative decisions. Additionally, it would allow an observer and display specific reference point to
be established in the pipeline (for example, the colorist and primary mastering display) which
imagery could be directly related to in the future when remastering content for different formats.
A physiologically based individual observer model is introduced by Asano and Fairchild1, as well
as a method for separating a population of observers into a limited number of observer categories,
each with their own associated CMF2. Building on this work, we present a metameric match
simulation pipeline that is computationaly simple and can be used practically in motion picture
color management. Using this pipeline and the categorical CMFs, we perform a simulation
experiment with real display spectra and natural images to observe the display-matching
variability which could occur among a population as a result of observer metamerism in a motion
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 3
picture viewing scenario. The results provide further evidence that inter-observer metameric
variability is a relevant problem in the context of natural images. Finally, we outline how this
pipeline can be used in motion picture color management.
Background
In the late 1920s William David Wright3 and John Guild4 each performed color matching
experiments where red, green, and blue lights were adjusted to match the single wavelength
output of a monochromator at increments across the visible spectrum of light. The responses of
observers were averaged, and the CIE 1931 2-degree field of view CMF was derived to achieve
a set of functionality objectives using this data as its basis (Figure 1a). The function has become
so deeply embedded in color engineering applications that to suggest its replacement or
ammendment on a large scale would likely result in compatibility problems which outweigh any
benefit to perceptual accuracy5,6. However, reliance on this average function to match between
displays of different spectral quality can lead to color management inconsistencies, as the
physiological CMFs of individuals can vary significantly from person to person. This variation is
the result of a wide range of factors, including genetics, age, and health & lifestyle factors like
whether the observer is a smoker, is overweight, has diabetes, etc. Together, these factors can
result in variation in the observer's optical pathway (density variation in the lens, macula, and
retinal photopigment) - altering the spectral characteristics of light absorbed by the
photoreceptors1.
a. b.
Figure 1a. CIE 1931 2-Degree field of view CMF () commonly used in colorimetric
measurements. b. Spectrally narrow band primary measurements for an IMAX laser projector.
While variation in these factors may only cause subtle or undetectable errors in cases where the
display primaries are spectrally broadband (Xenon lamp projectors, cathode ray tube displays,
film dyes, etc.), narrow-band primary displays like laser projectors (Figure 1b) can exacerbate
differences7. Unfortunately, these narrow band spectral architectures are required to meet the
high purity primaries specified in the current wide color gamut standard, Rec. 2020. In order to
assess the magnitude of these variations in a practical color grading scenario, Asano et al.
conducted a color matching experiment8 where 28 observers were asked to make matches
between images displayed on an LCD monitor and on a Pico laser projector using L*a*b* controls.
The results of several observers are shown in Figure 2a, compared to the match predicted using
the CIE 1964 10-degree standard observer. From a qualitative perspective, significant shifts in
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 4
color balance can be seen between observer responses, demonstrating that the color grading
tone and style of mastered images can be significantly impacted by metamerism error. To analyze
the results quantitatively, the group compared the intra to inter-observer variability and found that
each observer had an average match point that was significantly different from others.
a. b.
Figure 2. Results from Asano et al.8 a. Matching responses of several observers, compared side
by side providing a visualization of the potential variability which could result between observers
matching a laser projector and LCD screen and b. average responses for each individual observer
surrounded by their intra-observer error tolerances.
In order to simulate the color matching variance which could occur in a population of observers,
Asano and Fairchild developed a model which considers 10 parameters (age, visual degree
exposure to stimulus, lens pigment density, macular density, peak optical density for L, M, and S
photopigments, and deviations in peak sensitivities for the L, M, and S cones) which can be
measured from an observer's physiology1. Then, a series of studies were collected which reported
inter-observer variability in relation to these factors, and their results were analyzed to obtain
standard deviation values for each parameter. Then, a Monte-Carlo simulation generating 10,000
random color-normal observer CMFs considering these standard deviations was conducted.
Taking inspiration from the observer grouping experiment of Sarkar9, Asano and Fairchild show
that this large group of color matching functions can be simplified into a limited set of categorical
observers using a k-medoids clustering algorithm2. This technique is similar to k-means, however
its output is restricted to be a function within the set generated by the physiological model, such
that the categories produced could represent actual observers rather than functions derived as
an average between them. The technique of grouping observers into a small number of categories
allows for the discretization of population variability, while still making a significant improvement
in color matching prediction compared to the use of a single standard observer. This was tested
through a series of simulations matching a white patch rendered for different display
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 5
combinations. Using a large set of ground truth observers, a three-channel scaling factor was
determined for each observer to make a perfect match for every display combination. Then, the
match values for the test display were converted to CIE 1931 2-degree tristimulus values (XYZ)
via a 3x3 matrix. The values were then converted to CIELAB representation and their Delta E
2000 (ΔE00) difference to the nearest categorical observer’s match was taken. The mean ΔE00 of
the group of observers was then calculated, and the simulation was repeated with an increased
number of categorical observers (adding additional categories in order of their improvement to
overall error). The results are shown in Figure 3. Using a single observer category (equal to the
CIE perceptual observer given an observer age of 38 and a 2-degree field size), the highest
average error of 5 ΔE00 is obtained in the case of comparing a CCFL backlit LCD screen to a laser
projector. It can also be seen that the improvement in terms of average error slows significantly
at the 10-observer mark.
Figure 3. Average ΔE00 error between ground truth observer matches and the match generated
for the nearest categorical observer, as a function of the number of categorical observers allowed.
For the categorical observers to be used in a motion picture workflow, all individuals making color
critical decisions should be characterized. To accomplish this, several "observer calibration"
methods have been proposed. In general, these involve method of adjustments color matching
between color patches shown on spectrally dissimilar displays or are forced choice comparison
experiments between different observer category renderings. Sarkar proposes a practically useful
methodology9 where observers are presented a series of color patch pairs and are asked to report
the quality of the match between the two via a three-choice scale (not acceptable, acceptable,
satisfactory.) The color patch pairs are generated to each be a match for a specific observer type
given the displays used in the experiment. The results are tallied via a weighted scoring scheme
which puts a high penalty on unacceptable matches, a low reward for acceptable matches, and a
high reward for satisfactory matches. In extension of this work, a low cost hardware observer
calibrator is proposed10. Instead of using spectrally dissimilar displays to render color patches,
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 6
this device simply uses two sets of three narrow band LEDs which emit light at wavelengths where
the maximum error exists between observers.
Model
Given a categorical  consisting of (,,) and two three-channel sets of display
primary spectra (, , ), each sampled from 390-780nm at 5nm increments, we intend to find
the display linear  scalars to send to the second display  in order to match the  cone
excitations resulting from a given  triplet on the first display . The categorical s are
generated using the method and parameters from Table 1 of Asano and Fairchild2. No extra
normalization steps are applied to this data. The calibration of the display when the primary
spectra were measured is unknown, so we find scalars for each primary such that, at maximum
drive value, they produce the same pseudo-cone stimulus as prescribed for the white point of
Rec.709 for , which is the CIE 1931 2-degree observer.
The procedure for calibrating display is conducted in the following way:
 (1)
 (2)


 (3)
 
(4)
 (5)
Then, we carry normalization factors and a 3x1 calibration scalar vector 
 for both displays () to the matching step, along with , the
chosen categorical observer for the match:


 (6)
 
(7)
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 7
Now we have the forward transformation matrix  for  as seen by categorical observer
. Given the display linear  values of the image mastered on , this matrix will convert
to the  cone signal values as observed by . We now want to find the inverse
transformation matrix , which we can apply to the  values to find the RGB values to scale
 so  gets the same excitation while viewing . So, we repeat Equations 6 and 7 for
the case of  and apply  and  to the display linear input, .


 (8)
 
(9)
 (10)
Using this procedure, the linear representation of any image mastered on  can be transformed
to its linear representation for . For our purposes, we add an additional step to convert the 
output to CIE 1931 2-degree  such that we can pass our results to standard error metrics.
Starting with  (assuming that the CIE 1931 2-degree CMF was used as ) and display
, we scale each of the coefficients by its corresponding primary calibration value , giving
us the  values for each display primary at its maximum drive value.
 

 (11)
then, convert to chromaticity values.  (12)
Finally, we derive the primary matrix .
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 8
(13)



 (14)
 (15)
Finally, by applying  to , we get the CIE 1931 2-degree tristimulus values of the
observer match, .
 (16)
Simulation
In order to test the pipeline presented above, we performed a series of simulations using a
collection of six display spectra arranged in different combinations, a set of 41 images, and the
10 categorical observers proposed2, for which the
functions are shown in Figure 4. A subset
of these results (seven varied images from full test set) is shown in Table 1. Our display collection
is a subset of those tested in the experiment shown in Figure 3, and includes the following: Sony
BVM series CRT, Apple Cinema CCFL backlit LCD display, Samsung Galaxy S3 OLED,
Dreamcolor LED backlit LCD, IMAX laser projector, PTAX200U LCD projector. In this simulation,
for each image and display pair, we generate a match for every observer to be shown on the
second display in the pair to match the reference image shown on the first. We then take the error
between each observer match and a reference observer match (in this case we use the CIE 1931
2-degree CMF), simulating the range of error one could encounter within a population when
calibrating the two displays to a single reference observer. We also performed this experiment for
patches at the display maximum drive values and at white. The results are shown in Table 1 in
terms of three different error metrics. The first is the mean (across all pixels in the image, all
observers in the set) ΔE00. The second is the standard deviation in mean ΔE00 across the observer
set. Finally, we show the results of the recent observer metamerism metric ΔPOM211. This metric
is calculated as the number of pixels over a noticeable difference threshold (here we take ΔE00 =
2), which we convert to percentage of image area.
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 9
Figure 4. ,
, and functions for ten categorical observers from [2]
Table 1. Metamerism simulation results for a subset of images, display pairs. Top group: mean
ΔE00 error between observer categories and reference observer, averaged over image pixels and
categories. Center group: standard deviation between mean ΔE00 averaged over image pixels for
different observer categories. Bottom group: % image area over 2 ΔE00, averaged over observer
categories. The chart color coding scale is relative to the readings within the same black border
(per error metric, stimulus type.) Red boxes indicate a high relative error, followed by yellow and
green for lesser readings.
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 10
We can see from the results in Table 1 that the error values for the primary colors give a fairly
good indication of the amount of error we expect to see in the image. For example, there exists a
large degree of error between observers for blue compared to the other primaries, so images with
large blue regions tend to have a greater degree of metamerism error. We can also see from the
results of the bottom group that in many display pair/image cases, large regions of image area
would be detectably different between the categorical observers’ matches and that of a traditional
calibration pipeline.
Figure 5 shows an array of images matched for a given display pair and observer. These images
were generated first by transforming their display linear sRGB representations to CIE 1931 2-
degree , then transforming to the first display representation (in this case, the Sony BVM
series CRT) using  derived from Equations (11-15.) Then, the transform to the metameric
match for a given observer category shown on the destination display (IMAX laser projector) was
applied as in Equations (6-10). Finally, the images are converted back to  using , and
later back to the sRGB representation shown. This figure should be interpreted as a visual
example of the degree to which the color balance of an image may need to be shifted to make a
match on the same display pair for different observers, rather than a visualization of each
observer’s perception. This is an important distinction as this pipeline simulates only the initial
physiological stages of the visual pathway and does not consider any dynamic processing that
occurs between absorption and perception which could vary greatly between observers.
a. b.
c. d.
Figure 5. sRGB renderings of images shifted to make a metameric match between a Sony BVM-
CRT and IMAX laser projector for a set of observers . Panel a. shows the sRGB source image,
and the matches for observers 3 (panel b.), 7 (panel c.), and 10 (panel d.) are shown. Test image
provided by Arri Camera Systems.
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 11
Discussion
We propose the use of this model in a post-production color management pipeline in the following
way. First, the displays upon which color critical observers work will need to be calibrated to the
relevant mastering standard and characterized in terms of their EOTF and primary spectra at
maximum drive value. Then, a reference representation (display, observer) of the images can be
selected (the colorist and their primary reference monitor, for example.) Next, all color critical
observers will need to be tested for their categorical observer type. This can be done using the
method of Sarkar9. These experiments can be generalized to any display pair which is spectrally
incompatible enough to produce visible metamerism errors, so to perform the test, users can
select the two displays in their color management pipeline which are most likely to be a
problematic pair. The simulation procedure described above can give an indication whether
metamerism error is likely to be noticeable for a given display pair and image/color patch set.
For each destination display/observer pair, the process described in equations 1-10 can be
carried out, using the relevant calibration standard (observer, white point) in equations 1-5, the
reference display and observer data for Equations 6 and 7 to derive , and the destination
display and observer data for Equations 8 and 9 to derive . Then, taking the display linear
representation of the image mastered by the colorist on their reference monitor, we apply the 3x3
transformation , apply the inverse of the EOTF, and display the result. Using this
procedure, any color management inconsistency in the pipeline related to observer metamerism
is minimized. These transforms, however, should not be baked into distribution masters and
instead should only be used within the post production process to homogenize color rendering
between set workstations/observers. At the time of distribution, the color encoding of images
should align with the relevant distribution standard.
Before a recommendation for the use of this pipeline and categorical observer color matching
functions for post-production color management can be confidently made, we plan to perform a
psychophysical experiment to confirm that they consistently result in display matches for real
observers. In this experiment, observers will be presented with a series of images displayed side
by side on spectrally dissimilar displays and will be asked to rate the quality of the match on the
following scale (1- unacceptable match, 2 - acceptable match, 3 - satisfactory match.) The image
presentations will be repeated, applying transforms from the pipeline above such that they match
between the two displays for each observer category. Following this, the experiment will be
repeated using the same procedure but employing a set of color patches instead of natural
images. The results from each experiment will be tallied and compared between the natural image
and synthetic stimuli (color patch) tests. If the classifications from the natural image experiment
consistently match those of the synthetic stimuli experiment, we will have some confirmation that
the categorical observers allow for consistent color management improvements in the context of
post-production and natural images.
While the simulation results presented here indicate the metamerism error that could exist within
a population, given the images and displays tested, it will be important to have psychophysical
data with which to verify the use of these error metrics for natural images. To start, they rely on
the imperfect perceptual uniformity of the CIELAB space, such that equal error readings in white
and blue regions, for example, might correspond to different degrees of perceptual distance to
observers. In addition to this, our use of ΔE00 averaged over image pixels could be misleading as
there could be small image regions with very high error values which would skew the metric but
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 12
might not be noticed by observers. Our measurement of image area over a just noticeable
difference ΔE00 could also be skewed for the opposite reason - if one image contains large regions
that are just at the threshold, while another contains small but contextually significant regions that
have a very significant error. While the metrics are unaware of visual attention, one can assume
that the average viewer would detect a difference in two images if large portions of their image
area exceed a noticeable difference threshold, as was the case with many examples in our
simulation. It is possible that these selected metrics or others could be combined in some way for
a more accurate prediction of observer metamerism error in natural images, however this could
only be verified through psychophysical experimentation. For this reason, it would be highly
interesting to characterize the degree of perceptual error related to observer metamerism in still
and moving images in a post-production environment.
The big question with regard to the practical use of these profiles is always how they can be
leveraged for group viewing situations. If the pipeline above is applied directly to the source image,
the resulting rendering would only be correct within a degree of error for certain members of the
viewing group. To average between the two most prominent categories in the group would be
non-trivial, as it is unclear in what representation the operation would take place, or whether or
not the resulting CMF would correspond to the perception of any real observer. The experiments
of Sarkar9 showed that for his category set, there were a few sets of ‘twin’ categories for which
the matches were consistently similar, suggesting that rendering the image for one categorical
observer may not mean that the rendering would be erroneous for all other categories. However,
if we look at our pipeline, we see that the matching transformation is linear and is applied to
display linear luminance, so its possible that an optical filter could be created in order to transform
between a reference representation (CIE 1931 2-degree, for example) and each categorical
observer. Then, if the entire group is viewing the image through the filter corresponding to their
category, everyone will perceive the correct rendering of the image.
Conclusion
In this work, we present a computationally simple color management pipeline which accounts for
variation in observer physiology by incorporating the categorical observers of Asano and
Fairchild2. Using it, we perform a simulation experiment with real display spectra and natural
images to observe the rendering error which could occur among a diverse population as a result
of standard calibration routines. The results provide evidence that error related to this variability
is a relevant problem in the context of natural images. Finally, we outline how this pipeline can be
used in motion picture color management. For future work, a psychophysical experiment to verify
the accuracy of the pipeline and the consistency of the use of categorical observers in a natural
image viewing scenario will be conducted. Additional psychophysical experiments exploring the
perceived degree of metamerism error in natural images would also be of interest to perform.
Using this data, a more appropriate observer metamerism error metric based on image statistics,
spectral differences, or a combination of the two could be derived. A final useful advancement of
the work would be the development of simple online tools to assist practitioners in assessing the
potential variability of their display devices, calibrating color critical observers with regard to their
category, and implementing the pipeline in their color management ecosystem.
© 2020 Society of Motion Picture & Television Engineers® (SMPTE®) 13
Acknowledgements
This work has received funding from 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. PGC2018-099651-B-I00 (MCIU/AEI/FEDER, UE).
References
[1] Asano Y, Fairchild MD, Blondé L (2016) Individual Colorimetric Observer Model. PLoS
ONE 11(2): e0145671. doi:10.1371/journal.pone.0145671
[2] Yuta Asano, Mark Fairchild. (2020). "Categorical observers for metamerism." Wiley
Publishing. DOI: 10.1002/col.22493
[3] William David Wright (1928). "A re-determination of the trichromatic coefficients of the
spectral colors." Transactions of the Optical Society. 30 (4): 141164. doi:10.1088/1475-
4878/30/4/301.
[4] John Guild (1932). "The colorimetric properties of the spectrum." Philosophical Transactions
of the Royal Society of London. Series A, Containing Papers of a Mathematical or Physical
Character. 230: 149187.
[5] Thornton, W.A. (1992). Toward a more accurate and extensible colorimetry. Part Ii. Discussion.
Color Res. Appl., 17: 162-186. doi:10.1002/col.5080170305
[6] Hugh S. Fairman, Michael H. Brill, Henry Hemmendinger (1997). “How the CIE 1931 Color-
Matching Functions Were Derived from Wright-Guild Data.” Color Research and Application
Volume 22, Number 1 (pp. 11-23.)
[7] David L. Long and Mark D. Fairchild (2014). “Modeling Observer Variability and Metamerism
Failure in Electronic Color Displays.” 22nd Color and Imaging Conference.
[8] Yuta Asano, Mark D. Fairchild, Laurent Blondé, Patrick Morvan. (2014). "Observer Variability
in Color Image Matching on a LCD Monitor and a Laser Projector." 22nd Color and Imaging
Conference Proceedings.
[9] Abhijit Sarkar, Laurent Blondé, Patrick Le Callet, Florent Autrusseau, Patrick Morvan, Jürgen
Stauder. (2010). "Toward Reducing Observer Metamerism in Industrial Applications:
Colorimetric Observer Categories and Observer Classification." 18th Color and Imaging
Conference Proceedings.
[10] Patrick Morvan, Abhijit Sarkar, Jurgen Stauder, Laurent Blondé, Jonathan Kervec, Hasan
Sheikh Faridul. (2011). “A Handy Calibrator for Color Vision of a Human Observer.” ICME 2011,
Barcelona, Spain.
[11] Hao Xie, Susan P. Farnand, Micheal J. Murdoch. (2020) “Observer metamerism in
commercial displays.” Journal of the Optical Society of America. DOI: 10.1364/JOSAA382228
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
This study proposes a vision model for individual colorimetric observers. The proposed model can be beneficial in many color-critical applications such as color grading and soft proofing to assess ranges of color matches instead of a single average match. We extended the CIE 2006 physiological observer by adding eight additional physiological parameters to model individual color-normal observers. These eight parameters control lens pigment density, macular pigment density, optical densities of L-, M-, and S-cone photopigments, and λmax shifts of L-, M-, and S-cone photopigments. By identifying the variability of each physiological parameter, the model can simulate color matching functions among color-normal populations using Monte Carlo simulation. The variabilities of the eight parameters were identified through two steps. In the first step, extensive reviews of past studies were performed for each of the eight physiological parameters. In the second step, the obtained variabilities were scaled to fit a color matching dataset. The model was validated using three different datasets: traditional color matching, applied color matching, and Rayleigh matches.
Article
Full-text available
The variability among color-normal observers poses a challenge to modern display colorimetry because of their peaky primaries. But such devices also hold the key to a future solution to this issue. In this paper, we present a method for deriving seven distinct colorimetric observer categories, and also a method for classifying individual observers as belonging to one of these seven categories. Five representative L, M and S cone fundamentals (a total of 125 combinations) were derived through a cluster analysis on the combined set of 47-observer data from 1959 Stiles-Burch study, and 61 color matching functions derived from the CIE 2006 model corresponding to 20-80 age parameter range. From these, a reduced set of seven representative observers were derived through an iterative algorithm, using several predefined criteria on perceptual color differences (delta E*00) with respect to actual color matching functions of the 47 Stiles-Burch observers, computed for the 240 Colorchecker samples viewed under D65 illumination. Next, an observer classification method was implemented using two displays, one with broad-band primaries and the other with narrow-band primaries. In paired presentations on the two displays, eight color-matches corresponding to the CIE 10 ° standard observer and the seven observer categories were shown in random sequences. Thirty observers evaluated all eight versions of fifteen test colors. For majority of the observers, only one or two categories consistently produced either acceptable or satisfactory matches for all colors. The CIE 10 ° standard observer was never selected as the most preferred category for any observer, and for six observers, it was rejected as an unacceptable match for more than 50 % of the test colors. The results show that it is possible to effectively classify real, color-normal observers into a small number of categories, which in certain application contexts, can produce perceptibly better color matches for many observers compared to the matches predicted by the CIE 10° standard observer.
Article
A method is proposed to generate categorical colour observer functions (individual colour matching functions) for any field size based on the CIE 2006 system of physiological observer functions. The method combines proposed categorical observer techniques of Sarkar et al with a physiologically‐based individual observer model of Asano et al and a clustering technique to produce the optimal set of categorical observers. The number of required categorical observers varies depending on an application with as many as 50 required to predict individual observers' matches when a laser projector is viewed. However, 10 categorical observers are sufficient to represent colour‐normal populations for personalized colour imaging. The proposed and recommended categorical observers represent a robust and inclusive technique to examine and quantify observer metamerism in any application of colorimetry.
Article
Two uniform patches presented on two displays under identical viewing conditions can appear as the same color to one observer but as mismatched colors to another observer. This phenomenon, called observer metamerism (OM), occurs due to individual differences in color matching functions. To avoid its potentially adverse impacts in display calibration and characterization, it is desirable to have a predictive model of OM. In this work, we report the computational results of how to use existing metrics to quantify the potential OM between commercial display pairs and a proposed OM metric that is verified through a psychophysical experiment.
Article
Wide color gamut media are emerging in the market, and this trend has been accelerated by ITU recommendation, Rec.2020 in 2012. Wide color gamut media possess spectrally narrow primaries, which would potentially increase the degree of observer metamerism. In this study, it was investigated if observer metamerism could be a serious issue under practical viewing conditions. Namely, real images were used as a matching stimulus instead of uniform colors. We carried out the color image matching experiment on two different media: an Apple Cinema HD LCD monitor and a Microvision laser projector. The results from 28 color-normal observers were analyzed. The obtained inter-observer variability was large enough that observer metamerism would be a serious issue where the laser projector is viewed together with conventional media. Each observer had a match point that was significantly different from those of other observers. It was found that effective field size changes (and an observers CMFs change) depending on image contents. Complex images require smaller field size whereas simple images require larger field size.
Article
The electronic display industry has begun a migration towards higher color gamut devices driven by LED, OLED, quantum dot and laser technologies capable of generating near monochromatic color stimuli in the traditional red, green, blue three-channel paradigm. The use of highly selective spectral stimuli, however, poses a risk to the consistency of visual experience amongst a group of disparate, but otherwise normal, color observers. Several models of spectral color vision have surfaced in recent research and are helping investigators to better understand the implications for color experience variability. The present research serves to summarize various color difference indices that may be useful in predicting the magnitude of observer response inconsistencies and applies them to simulations of current electronic displays as examples of potential concerns these new high-gamut technologies might raise. In particular, various laser-based displays are shown to perform with significantly increased observer variability versus traditional ITU-R Rec. 709 and SMPTE 431 RGB-primary displays utilized in the cinema industry. Further, observer metamerism can be reduced significantly with proper optimization of a multichannel projection system comprising seven explicitly designed primary spectra.
Article
The principles that guided the founders of the CIE 1931 system for colorimetry are examined. The principles are applied to the Wright-Guild experimental determinations of the color mixture data to show in detail how and why each step in the development of the CIE 1931 system for colorimetry came about. These steps are examined in the light of 65 years advanced knowledge of colorimetry. The necessity for each of these principles in the modern world is examined critically to determine whether one might hold to the same principles if the system were being freshly formulated today. © 1997 John Wiley & Sons, Inc. Col Res Appl, 22, 11–23, 1997.
Article
Using a new trichromatic colorimeter a series of colour matches through the spectrum has been made by ten observers. The results have been averaged and a mean set of trichromatic coefficients for the spectral colours derived. These results are compared with previous determinations made by König and Abney. The variations in the coefficients that have been found amongst the ten observers must, as a consequence of a new method of basing the trichromatic units, be attributed to variations in the process of reception, but their magnitude appears to be of a small order. On the other hand, there are big differences in the amount of the macular pigment in different eyes and probably some variation in its dominant hue. These variations have been investigated by matches on a standard white, results for 36 observers being given in the paper and a mean value determined. This value, combined with the mean spectral coefficients, has been used to compute an average locus for the spectral colours in the colour triangle, with white at the centre. Other points discussed in the paper include the technique of colour matching, the range of intensity over which matches remained valid, and variations of luminosity.
Article
Technical advances in the forty years since the Stiles-Burch color-matching work-among them the bright split-field visual colorimeter coupled by quartz light pipe to a spectroradiometer-allow measurement of an accurate absolute spectral power distribution (SPD) of every viewed light. Furthermore, the ability to measure accurately the power-content of each spectral constituent of a viewed light has clarified both colorimetric procedures and signficance, for example, of color-matching functions (CMFs) and tristimulus values. Three disparate sets of three spectral primaries were used throughout the work. This has not been done before, and its use elucidated much of the difficulty (as described by Judd, Stiles, MacAdam, Wyszecki, and others) inherent in traditional colorimetry. Errors in computed chromaticities of sets of 28 visually-matching lights depend strongly on their spectral content, and are particularly large in the presence of high spectral content in the violet, deep red, and near 500 nm or 580 nm. Peaks of visual efficiency, and minima in chromaticity errors, were found with power in three intermediate spectral regions near 450 nm, 530 nm, and 610 nm. Computed luminance (1964 CIE Standard Observer) and power-content also depend strongly on spectral content of the visually-matching white lights. That this is true of luminance is generally unknown. Perceived brightness of single spectral lights in isolation, and of mixtures, was examined at length: As a rule, removing one spectral component from a mixture increases perceived brightness, thus invalidating both luminance and R + G + B as correlates of brightness. Visual tests of the Grassmann proportionality and additivity assumptions, with highly metameric mixtures, tended to confirm them. Nevertheless, numerous tests-substituting pairs of actual visually-matching lights from one primary-set for pairs from another, which is how “transformation of primaries” is (mathematically) carried out-led to the inescapable conclusion that the normal human visual system does not honor such transformation. At the same time, numerous indications supported a particular set of primaries as an invariant of the visual system, making transformation needless as well as invalid. Chromaticity errors are not the fault of the particular maximum-saturation CMFs utilized; the observer's own CMFs do no better, whatever the primary-set. Soundness of traditional construction of the chromaticity diagram is shown to depend on the primary-set, breakdown occurring for some otherwise legitimate primary-sets.