Ciencia y La Tecnologia Del Color Español

CIENCIA Y TECNOLOGÍA
DEL COLOR
2008 WORKSHOP ON
“COLORIMETRY AND COLOR IMAGING”
CIENCIA Y TECNOLOGÍA
DEL COLOR
2008 WORKSHOP ON
“COLORIMETRY AND COLOR IMAGING”
Red temática “Ciencia y Tecnología del Color”. Workshop on “Colorimetry and Color Imaging”
Título: Ciencia y Tecnología del Color 2008. Workshop on “Colorimetry and

Color Imaging”
Editores: Joaquín Campos Acosta y Rafael Huertas Roa, Red Temática

“Ciencia y Tecnología del Color”
Depósito Legal: GR-4028/2009
ISBN: 978-84-92680-76-4
Edita e Imprime: Copicentro Granada S.L.
24-26 de Septiembre de 2008

Almuñécar (Granada)
INTRODUCCIÓN
En 2008 la reunión anual de la red “Ciencia y Tecnología del Color” ha adoptado la

fórmula de workshop, bajo el título “Colorimetry and Color Imaging”. La razón que nos
movió a organizar dicho evento fue la constatación de que en la sociedad actual son
muchos los problemas científico-técnicos asociados a la medida y reproducción del color
de los productos naturales o manufacturados. El mercado exige como norma de calidad que
el color de cada producto lo caracterice de forma única y que sea el esperado por el
consumidor.
Añadido a ello, en los últimos tiempos, el desarrollo de los dispositivos multimedia ha
dado una dimensión nueva al control del color en imágenes. Nos referimos a la generación,
procesado y reproducción del color en dispositivos como monitores (CRT, LCD), scanners,
cámaras digitales o impresoras. Ello ha dado lugar a nuevas necesidades en el campo de las
tecnologías asociadas a las imágenes en color que superan ampliamente los que se tenían
en la fotografía convencional o la impresión gráfica clásica. En este campo se trabaja
intensamente en todo el mundo, tanto desde el punto de vista de la investigación básica
como aplicada.
En el taller nos propusimos tanto suministrar formación en aspectos esenciales del mundo
de la Colorimetría y de la adquisición, procesado y reproducción de imágenes en color,
como poner en contacto los distintos sectores involucrados en este mundo. Nos referimos
al mundo de la universidad, del CSIC y de la industria. Por ello invitamos a participar en el
mismo a investigadores consagrados y en formación, a tecnólogos y a gestores de los
distintos ámbitos. Con ello creamos un foro de intercambio de opiniones alrededor de
temas concretos relacionados con el color.
Los temas tratados en este taller fueron los de más interés en el momento actual:
diferencias y tolerancias de color, modelos de apariencia de color, evaluación colorimétrica
de imágenes en color, adquisición, procesado y reproducción de color. Para ello contamos
con especialistas españoles y extranjeros de máximo nivel, y encontramos un auditorio en
el que se conjugaron a su vez especialistas en estas materias con personas en formación.
Queremos destacar la participación de los 6 conferenciantes. Los profesores Regino Saéz y
Eduardo Gilabert abordaron temas relacionados con la Química del color y los nuevos
pigmentos ecológicos. El profesor Changjun Li abordó las recomendaciones recientes en el
cálculo de valores triestímulo. Por otra parte, los profesores Marku Hauta-Kasari, Alain
Tremeau y Jaume Pujol, abordaron aspectos relacionados con la adquisición, procesado y
evaluación de imágenes espectrales y en color. Las conferencias fueron completadas con la
presentación por parte del Dr. David Roldán de la empresa Aquateknica de software para
la formulación de color, de una mesa redonda sobre el futuro de las investigaciones en
estos campos y de la tradicional exposición de trabajos por parte de estudiantes de
doctorado, nueve en total, que presentaron sus trabajos de forma muy correcta y eficiente.
Dichos trabajos son los que se presentan en los resúmenes de esta publicación.
En nombre del comité organizador:
Javier Romero Mora.
Dpto. de Óptica. Universidad de Granada.

Índice

1. Evaluating a colour difference formula in a new multistage colour vision model. E.

Chorro, E. Perales, F. M. Martínez-Verdú, D. de Fez, M. J. Luque, P. Capilla .................. 9
2. Analysis of the modulation transfer function (MTF) spectral behaviour in detector

arrays. A. Fernández-Oliveras, A. M. Pozo, M. Rubiño ..................................................... 13
3. Color in virgin-olive-oil tasting glasses. L. Gómez-Robledo, M. Melgosa, F. J. Heredia,

M. J. Moyano, R. Roa, R. Huertas ....................................................................................... 20
4. Iris color evaluation through multispectral systems. J. Herrera, M. Vilaseca, J. Pujol,

M. Arjona, M. de Lasarte .................................................................................................... 23
5. Reproducción espectral de valores triestímulo mediante descripciones B-Spline:

evaluación del error en el color. C. Pizarro, J. Arasa, M. de Lasarte, J. Pujol, M. Arjona,
M. Vilaseca .......................................................................................................................... 28
6. Colorimetric properties of thermoplastic polymers coloured with nanopigments and

conventional pigments. V. Marchante, F. Martínez-Verdú, A. Marcilla, M. Beltrán ......... 33
7. The use of optimal stimuli in colour appearance modelling. E. Perales, C. Li, E. Chorro,
V. Viqueira, F. Martínez-Verdú, M. R. Luo ......................................................................... 37
8. Spectral and photometric analysis of textured surfaces. C. Plata, J. L. Nieves, E. M.

Valero, J. Romero ................................................................................................................ 42
9. ¿Son temporalmente estables las fuentes luminosas de las cabinas de iluminación? R.

Roa, R. Huertas, L. Gómez-Robledo, M. Melgosa .............................................................. 46

Evaluating a colour difference formula

in a new multistage colour vision model
E. Chorro (1), E. Perales (1), F. M. Martínez-Verdú (1),

D. de Fez (1), M. J. Luque (2), P. Capilla (2)
1. Departamento de Óptica, Farmacología y Anatomía. Universidad de Alicante.

Campus de San Vicente del Raspeig, Ap. Correo 99, E-03680 Alicante.
2. Facultad de Física. Universidad de Valencia. C/Doctor Moliner nº 50, 46100 Burjassot (Valencia).
E-mail: elisabet.chorro@ua.es
ABSTRACT:
The main purpose of this research line is to determine a methodology in order to develop a colour
vision model with a colour difference formula. In this review, we summarize the previous
research as well as the further research and the results that we expect to find.
Firstly, we describe briefly the chosen neural model, we study the uniformity of its perceptual
space, we propose a colour difference formula for it and finally we compare the predicion’s
model with those of the other models.
Considering the results, we will propose some improvements in the model, which will be
accounted in the further research.
Keywords: Colour appearance model, colour difference formula, neural model.
REFERENCES AND LINKS

[1] J. Gómez-Chova, P. Capilla, J. M. Artigas, M. J. Luque, A. Felipe, “ATTD: a new colour vision model
based on the physiology of the visual system”, in Proceeding of the 10th Congress of the International
Colour Association (AIC’05) May 8-13, Granada, Spain, pp. 1007-1010 (2005).
[2] J. Gómez-Chova, J. M. Artigas, P. Capilla, A. Felipe, M. J. Luque, “Testing the performance of ATTD
colour vision model”, in Proceeding of the 10th Congress of the International Colour Association (AIC’05)
May 8-13, Granada, Spain, pp. 1011-1014 (2005).
[3] CIE Publ. No. 159:2004, A Colour Appearance Model for Colour Management Systems: CIECAM02, CIE
Central Bureau (2004).
[4] E. Chorro, F. M. Martínez-Verdú, D. Fez, P. Capilla, M. J. Luque, “Analysing the metrics of the perceptual
space in a new multistage physiological colour vision model”, Color Res Appl. Accepted 3 July 2008.
[5] S. Guan, M. R. Luo, “Investigation of Parametric Effects Using Small Colour Differences”, Color Res
Appl. 24, 331-343 (1999).
[6] P. A. García, R. Huertas, M. Melgosa, G. Cui, “Measurement of the relationship between perceived and
computed color differences”, J Opt Soc Am A. 24, 1823-1829 (2007).
[7] M. Melgosa, R. Huertas, M. J. Rivas, P. A. García, A. González, M. Vik, G. Cui, “Significación estadística
de las correcciones introducidas en las fórmulas de diferencia de color CIE94 y CIEDE2000”, in
Proceeding of the VIII Congreso Nacional de Color 2007, September 19-21, Madrid, Spain (2007).

9 Almuñécar (Granada)
[8] M. R. Luo, G. Cui, B. Rigg, “The Development of the CIE 2000 Colour-Difference Formula: CIEDE2000”,
Color Res Appl. 26, 340-350 (2001).
[9] M. R. Luo, C. Li, G. Cui, “Combining Colour Appearance model with Colour Difference Formula”, in
Proceeding of the 10th Congress of the International Colour Association (AIC’05) May 8-13, Granada,
Spain, pp. 235-240 (2005).
[10] M. R. Luo, G. Cui, C. Li, “Uniform Colour Spaces Based on CIECAM02 Colour Appearance Model”,
Color Res Appl. 31, 320-330 (2006).
[11] S. S. Guan, M. R. Luo, “A colour-difference formula for assessing large colour differences”, Color Res
Appl. 24, 344-355 (1999).
[12] A. Valberg, Light Vision Color, Editorial Wiley, England (2005).
[13] R. DeValois, K. K. DeValois, E. Switkes, L. Mahon, “Hue Scaling of Isoluminant and Cone-spedific
Lights”, Vision Res. 37, 885-897 (1997).
1. Introduction after a last transformation, yield the achromatic, red-

The main purpose of this research line is to green, and blue-yellow perceptual mechanisms. And
determine a methodology in order to develop our the perceptual descriptors can be calculated from the
colour vision model with a colour difference perceptual mechanisms. In the figure 1 you can see a
formula. flowchart of the ATTD05 model.
The multistage colour vision model, ATTD051-

2
, is the model that we have used up to now, because
this model is a neural model, and particularity, this
model tries to follow all stages of human visual
system. Other models3 can predict perceptual
attributes or some appearance descriptors too, but
the chosen model is able to predict perceptual
descriptors and some appearance effects of human
visual system in a natural way. Moreover if we can
supply it with a colour difference formula, the final
result will be a potent tool for a lot of aspects related
to colour technology and science (appearance,
reproduction, modelling, etc).
2. Previous Research
Our start point was the ATTD05 model, which is a
neural model that can predict the perceptual
descriptors, but initially this model is not useful for
predicting colour differences.
2.a. Description of the ATTD05 model
The ATTD05 model1-2 has three opponent
transformations, one at the retinal-LGN level and
two at the cortical level, multiplicative and
subtractive adaptation mechanisms and non-
Fig.1. Flowchart of the ATTD05 model.
linearities of the Naka-Rushton type.
Basically, the cone responses are the result of a
Thirty free parameters were determined by
first linear stage and after another lineal combination
minimizing the average distances between the
yields the responses of an achromatic channel, two
predictions of the model and a set of training
opponent channels with red-green opponency but
experimental data. The training data set included the
different polarities and one opponent channel with
spectral sensitivity of LGN opponent and non-
blue-yellow opponency. Finally the channels ATD,
opponent cells, as well as those of the chromatic and

achromatic perceptual mechanisms derived from hue experimental data set was the same data used for the
cancellation experiments in both neutral and development of CIEDE20008 and included 3657
chromatic adaptation conditions. But no data of the sample pairs with an average of 2.6 ΔE units.
colour difference were used to determine the
We found that Colour Difference Formula’s
parameters of the model.
predictions from ATTD05 were better than the
2.b. Uniformity of the ATTD05 model predictions from CIELAB for Sort Colour
Differences and better that CIELAB, CIE94 and
Next, we studied the uniformity of the
CIEDE2000 for Long Colour Difference data set,
perceptual space of the ATTD05 model4, because if
but were worse than the predictions from
the model yielded a uniform perceptual space we
CIECAM029-11.
could propose the colour difference formula as a
Euclidian distance between two points.
3. Further Research
From now on, our aim is to improve the obtained
results, and for this reason, we have proposed to
improve some stages of the model, because if the
model is better, the obtained results with it will be
better too. But restructuring the stages implies that
we have to calculate the coefficients and test the
model again. When we have a better model, we will
implement Colour Difference Formula as in the
previous way.
For this reason the future research will have to
be structured in the next way:
3.a. Restructuration of the ATTD05 model
In order to improve the model’s predictions we
propose some changes into the model, the most
Fig.2. Samples of constant Munsell Value 5 plotted on the important change is in the pre-cortical stage.
perceptual plane of the ATTD05 model.
At the first version of the model we calculated
two opponent channels with red-green opponency
In the figure 2 you can see the samples of the but different polarities because we assumed that
Munsell with a constant value in the perceptual these channels would be supported by Parvo cells
space of ATTD05. If the samples appeared with Type I receptive fields, with polarity +L-M
uniformly spaced in circles was because the mediated by L-on centre cells and polarity +M-L
perceptual space was uniform and, as you can see, mediated by M-on centre cells, and we supposed that
the space was quite uniform. But this was only a existed their symmetric receptive fields, mediated by
qualitative evaluation. Others quantitative L-off and M-off centre cells. In the new version we
parameters were calculated for study of the suppose that the responses of on-centre and off-
uniformity, as the position and eccentricity of the centre cells are not symmetric12-13, and therefore our
centre of gravity or circularity of the rings. With proposed model will have four opponent channels in
these parameters we could conclude that the this stage.
ATTD05 was uniform and therefore a colour Because the model will be restructured, the
difference could be computed as the distance first step will be to calculate the coefficients of the
between two points, with a weighing Euclidean new version of the model. Again the free parameters
metric, that is, with weight factors in each will be determined by minimizing the average
coordinate, so we had to fit experimental data set in distances between the predictions of the model and a
order to obtain the coefficients. training experimental data set. But now, the training
2.c. ATTD05 colour difference formula versus data set will include the spectral sensitivity of LGN
other models opponent and non-opponent cells, as well as those of
the chromatic and achromatic perceptual
Finally, we compared predictions4 made with mechanisms derived from hue cancellation
our colour difference formula and with other models experiments of DeValois and DeValois13.
using statistical test based on parameters as
Performance Factor5 and STRESS6-7. The

3.b. Study of the Uniformity and comparison with

the previous one
We will have to test the results of the model
checking that the model reproduces some colour
appearance effects or the results are fitted to
experimental data set.
Again, after checking the uniformity of the
perceptual space we will propose a colour difference
formula as the distance between two points, with a
diagonal metric, so we will have to fit experimental
data set in order to obtain the coefficients. These
experimental data sets will be the data used for the
development of CIEDE2000, again.
3.c. Propose a colour difference formula and test
it
Finally, if obtained results are the expected
results, we will propose a colour difference formula,
as a Euclidean distance with a weight factor and we
will compare predictions made with our colour
difference formula and with other models using
statistical test based on parameters as Performance
Factor and STRESS.
4. Expected Results
Our first change in the model has been proposed in
order to do matching the model with the visual
system and for this reason we expect that the stages
of the new model will be more similar than in earlier
version to the stages of the visual system. On the
other hand if the model has improved, the uniformity
of the model should be improved too. And therefore
the Colour Difference Formula’s predictions from
new model should be better than the predictions
from earlier version of the model.
Acknowledgements
This research was supported by the Spanish Ministry
for Education and Science by means of the grant
number DPI2005-08999-C02-02.

Analysis of the modulation transfer function (MTF) spectral

behaviour in detector arrays
A. Fernández-Oliveras, A. M. Pozo, M. Rubiño
Departamento de Óptica. Facultad de Ciencias. Universidad de Granada.

Campus Fuentenueva s/n, 18071 Granada.
Email: alilia@ugr.es
ABSTRACT:
Today, CCD and CMOS detector arrays present excellent features in imaging systems. To
investigate the suitability of each technology according to the application, in this work, we have
used the speckle method to determine the modulation transfer function (MTF) at different
wavelengths of the visible spectrum, for the detectors of two scientific cameras (a CCD and a
CMOS). For the CMOS detector, the differences between the MTF curves intensify as the spatial
frequency augments while the MTF decreases as the wavelength increases. For the CCD detector,
the MTF spectral behaviour does not show this trend and the differences between the MTF curves
corresponding to extreme wavelengths are not expected to be significant.
Keywords: CCD and CMOS detectors, spectral image quality, modulation transfer function
(MTF), MTF spectral variation, speckle method.

[1] G. C. Holst and T. S. Lomheim, CMOS/CCD Sensors and Camera Systems, JCD Publishing, Winter Park,
FL. and SPIE Optical Engineering Press, Bellingham, WA, (2007).
[2] H. Helmers and M. Schellenberg, “CMOS vs. CCD sensors in speckle interferometry”, Opt. Laser
Technol., 35, 587–595 (2003).
[3] D. Litwiller, “CCD vs. CMOS: facts and fictions”, Photonics Spectra, 154–158 (2001).
[4] J. Janesick, “Dueling detectors. CMOS or CCD?”, SPIE’s OE Magazine, 41, 30-33 (2002).
[5] J. Janesick, “Lux transfer: complementary metal oxide semiconductors versus charge-coupled devices”,
Opt. Eng., 41, 1203–1215 (2002).
[6[ G. Deptuch, A. Besson, P. Rehak, M. Szelezniak, J. Wall, M. Winter and Y. Zhu, “Direct electron imaging
in electron microscopy with monolithic active pixel sensors”, Ultramicroscopy, 107, 674–684 (2007).
[7] S. K. Park, R. Schowengerdt and M. Kaczynski, “Modulation-transfer-function analysis for sampled image
system”, Appl. Opt., 23, 2572-2582 (1984).
[8] J. C. Feltz and M. A. Karim, “Modulation transfer function of charge-coupled devices”, Appl. Opt., 29,
717-722 (1990).
[9] D. N. Sitter Jr., J. S. Goddard, and R. K. Ferrell, “Method for the measurement of the modulation transfer
function of sampled imaging systems from bar-target patterns”, Appl. Opt., 34, 746-751 (1995).
[10] A. Daniels, G. D. Boreman, A. D. Ducharme, and E. Sapir, “Random transparency targets for modulation
transfer function measurement in the visible and infrared regions”, Opt. Eng., 34, 860-868 (1995).
[11] S. M. Backman, A. J. Makynen, T. T. Kolehmainen, and K. M. Ojala, “Random target method for fast MTF
inspection”, Opt. Express, 12, 2610-2615 (2004).

[12]. N. Guérineau, J. Primot, M. Tauvy and M. Caes, “Modulation transfer function measurement of an infrared
focal plane array by use of the self-imaging property of a canted periodic target”, Appl. Opt., 38, 631-637
(1999).
[13]. M. Marchywka and D. G. Socker, “Modulation transfer function measurement techniques for small-pixel
detectors”, Appl. Opt., 31, 7198-7213 (1992).
[14] J. E. Greivenkamp and A. E. Lowman, “Modulation transfer function measurements of sparse-array sensors
using a self-calibrating fringe pattern”, Appl. Opt., 33, 5029-5036 (1994).
[15] A. M. Pozo, A. Ferrero, M. Rubiño, J. Campos and A. Pons, “Improvements for determining the
modulation transfer function of charge-coupled devices by the speckle method”, Opt. Express, 14, 5928-
5936 (2006).
[16] G. D. Boreman and E. L. Dereniak, “Method for measuring modulation transfer function of charge-coupled
devices using laser speckle”, Opt. Eng., 25, 148-150 (1986).
[17] A. D. Ducharme, "Microlens diffusers for efficient laser speckle generation", Opt. Express, 15, 14573-
14579 (2007).
[18] G. D. Boreman, Y. Sun and A. B. James, “Generation of laser speckle with an integrating sphere”, Opt.
Eng., 29, 339-342 (1990).
[19] M. Sensiper, G. D. Boreman, A. D. Ducharme and D. R. Snyder, “Modulation transfer function testing of
detector arrays using narrow-band laser speckle”, Opt. Eng., 32, 395-400 (1993).
[20] A. M. Pozo and M. Rubiño, “Comparative analysis of techniques for measuring the modulation transfer
functions of charge-coupled devices based on the generation of laser speckle”, Appl. Opt., 44, 1543-1547
(2005).
[21] J. W. Goodman, Laser Spekle and Related Phenomena, J. C. Dainty, Ed., Vol. 9 of Topics in Applied
Physics Springer-Verlag, Berlin, pp. 35-40 (1984).
[22] L. I. Goldfischer, “Autocorrelation function and power spectral density of laser-produced speckle patterns”,
J. Opt. Soc. Am., 55, 247-253 (1965).
[23] G. D. Boreman, “Fourier spectrum techniques for characterization of spatial noise in imaging arrays”, Opt.
Eng., 26, 985-991 (1987).
[24] A. Fernández-Oliveras, A. M. Pozo and M. Rubiño, “Análisis de la variación espectral de la MTF de
videocámaras CCD mediante patrones de moteado láser”, Opt. Pura Apl., 41, 221-225 (2008). Article in
press.
[25] A. Fernández-Oliveras, A. M. Pozo and M. Rubiño, MTF spectral-variation comparison of detector arrays
used in multispectral imaging systems by speckle patterns, in Proceeding of the IS&T’s 4th European
Conference on Colour in Graphics, Imaging, and Vision. 10th International Symposium on Multispectral
Colour Science, IS&T, Springfield, VA, pp. 461-466 (2008).
[26] G. C. Holst, CCD Arrays, Cameras and Displays, JCD Publishing, Winter Park, FL. and SPIE Optical
Engineering Press, Bellingham, WA, (1996).
[27] I. Shcherback and O. Yadid-Pecht, CMOS Imagers: From Phototransduction to Image Processing, O.
Yadid-Pecht and R. Etienne Cummings, Eds., Kluwer Academic, Norwell, MA, pp. 53-74 (2004).
1. Introduction Therefore, to investigate the suitability of the

Cameras with imaging devices based on CCD and use of one or the other technology according to the
CMOS detector matrices1 are being used more and specific application of the camera, the complete
more in such disparate scientific and technological characterization of the different types of detector
fields as Colorimetry, Illumination, and matrices becomes necessary.
Astrophysics. Currently, both types of devices offer A system is optically characterized by the
excellent features in imaging systems when they are modulation transfer function (MTF), the
appropriately designed. The consensus is that the determination of which enables the image produced
two technologies complement each other and will by the system to be evaluated from its response in
coexist in the future, depending on the application spatial frequency7,8.
involved2-6.

For measuring the MTF of solid-state cameras, PSDoutput (ξ ,η ) = [MTF (ξ ,η )]2 PSDinput (ξ ,η ) (1)
the literature cites different methods that differ
essentially in the type of target or pattern used as the where ξ and η are the spatial frequencies
object. Thus, for example, methods use bar targets9, corresponding to the horizontal and vertical
random targets10,11, canted self-imaging targets12, directions x and y, respectively.
interferometric fringes13,14. PSDoutput is determined from the speckle pattern
One of the methods to measure the MTF, captured with the detector, being proportional to the
established in our laboratory, is based on using a squared magnitude of the Fourier transform of this
laser speckle pattern as the object15-20. This method speckle pattern. In the case of a rectangular single-
is suitable for analysing the detector independently slit, PSDinput is given by21,22:
of the camera lens, given that it does not require a
PSDinput ( ξ ,η ) = I
2⎡
⎢δ ( ξ ,η ) +
(λz )2 tri⎛⎜ λz ξ ⎞⎟tri⎛⎜ λz η ⎞⎟⎤⎥ (2)
lens to project the pattern. Furthermore, using a ⎜l ⎟ ⎜l ⎟
⎢⎣ l1l2 ⎝ 1 ⎠ ⎝ 2 ⎠⎥⎦
tunable laser source, we can characterize the device
at different wavelengths, this proving indispensable where| tri( X ) = 1 − X for X ≤ 1 and zero elsewhere;
< I > 2 is the square of the average speckle
in multispectral and colour-measuring applications.
irradiance; δ(ξ,η) is a delta function; l1 and l2 are,
Speckle is an interference phenomenon that respectively, horizontal and vertical dimensions of
occurs when coherent radiation is scattered from a the single-slit; λ is the wavelength of the laser; and z
rough surface. Several techniques can be used to is the distance between the single-slit aperture and
generate the speckle pattern, such as different types the detector.
of transmissive diffusers (ground glass15, fused
silica16, microlens arrays17) or integrating spheres18- Given the geometry of the single-slit, the
20
. PSDinput can be separated into frequencies ξ and η
the horizontal PSDinput(ξ,η) is the η=0 profile of
In the former case, an aperture situated in front PSDinput(ξ,η). This means that the MTF can be
of the integrating sphere enables us to specify the determined separately for x and y directions. In the
content of spatial frequencies of the speckle pattern. present work, we determine the horizontal MTF.
Two of the apertures used to date are the single-slit18 This can be done in a similar way for the vertical
and double-slit19, both of which present advantages direction.
and drawbacks20. In this work, we have used a
single-slit situated at the exit port of an integrating
sphere. 3. Method
It bears noting that in the works cited above, 3.a. Experimental set-up
the systems analysed are based generally on
Figure 1 presents the experimental set-up used.
scientific CCD cameras, and comparisons were not
It is composed of a tunable argon-ion laser source
made between devices of different technology, nor
(130 mW) or a He-Ne laser source (λ=632.8 nm; 17
was the MTF spectral variation studied.
mW), depending on the wavelength which is being
The aim of the present work is to apply the studied, an integrating sphere to generate the speckle
optical detector-characterization method, based on pattern (inner diameter of 152.4 mm), a polarizer to
the measurement of the MTF with speckle patterns, provide a linearly polarized laser-speckle pattern, a
to the analysis at different wavelengths of the image single-slit (6 mm height and variable width), and an
quality provided by different cameras. optical bench to hold the detector, which is
connected to the control card installed in a personal
For this, we have comparatively studied the
computer.
resulting MTF curves at different wavelengths of the
visible spectrum, for the detectors of two scientific The laser radiation is aimed at the entrance port
cameras (a CCD and a CMOS). of the integrating sphere, generating the speckle
pattern at the exit port. The aperture situated at the
exit port of the sphere (single-slit) determines the
2. Theoretical background content in spatial frequency of the pattern registered
The relationship between the theoretic power- in the detector. Under these conditions, the linear
spectral density known for a single-slit (PSDinput) polarizer ensures that the PSDinput is given by Eq.
and the measured power-spectral density (PSDoutput) (2)22.
allows us to determine the MTF of the detector by
means of the expression18:

direction, corresponding to a Nyquist frequency of

100 cycles/mm given by Eq. (4).
The width of the single-slit used was l1=3 mm
for the two scientific detectors (the CCD and the
CMOS).
Taking into account the width of the single-slit
and Nyquist frequency, we can calculate the distance
z between the detector and the aperture by using Eq.
(3) for each wavelength studied. The corresponding
values are listed in the following tables for each
detector analysed.
Fig. 1. Experimental set-up for the measurement of the
MTF of the detectors. The aperture A (single-slit) and the
polarizer P are situated at the exit port of the integrating TABLA I
sphere.
Distance between the detector and the single-slit aperture
for scientific camera CCD detector.
With the single-slit, the MTF can be Distance detector-
determined from a single measurement without the Wavelength (nm)
aperture (mm)
need to move the detector, but it must be situated at 632.8 44
a distance from the aperture in such a way that the 514 54
maximum input spatial frequency is equal to the 488 57
Nyquist frequency of the detector16,18. In this way, 457 61
the MTF can be determined in the largest possible
frequency range, and thus aliasing is avoided.
The distance z between the detector and the TABLA II
single-slit aperture can be calculated by the Distance between the detector and the single-slit aperture
expression: for scientific camera CMOS detector.
l1 Distance detector-
z= (3) Wavelength (nm)
aperture (mm)
λ ξ Ny 632.8 47
514 58
where l1 is the slit width, λ the wavelength of the 488 61
laser, and ξNy is the Nyquist spatial frequency of the 457 66
detector in the horizontal direction. For a detector
array with a centre-to-centre spacing between the
photoelements Δx, the Nyquist frequency is given 3.b. Data processing
by:
Once the detector was set at the corresponding
1 distance from the single-slit aperture, as indicated in
ξ Ny = (4) the previous section, the PSDoutput(ξ) was
2Δx
determined in the following way:
In this work, measurements were made using
For a given digitized frame of speckle data, a
the detectors of two different scientific cameras: a
region of 500x500 pixels was selected. Each
CCD and a CMOS.
horizontal row of data is a single observation of an
The scientific CCD camera had a high- ergodic random process. A fast Fourier transform
resolution CCD B/W PixelFly array of 1360x1024 (FFT), which is a discrete Fourier transform, was
pixels with a centre-to-centre spacing between them performed on each row of speckle data. The
of 4.65 µm. Consequently, with Eq. (4) taken into magnitude squared in one dimension provided a
account, the Nyquist frequency of this detector is single estimate of the one-dimensional power
107.53 cycles/mm in both directions. spectrum, PSDoutput(ξ). These 500 spectra were
averaged, for a better signal-to-noise ratio in the
The CMOS camera used was a CMOS B/W
PSDoutput(ξ)23. To reduce the noise even further, the
Atmos Areascan 1M30, the detector array of which
average was taken for 10 frames.
had 1312x1024 pixels. In this case the pixel pitch
was 5 µm in the horizontal as well as in the vertical

The frames were stored in tiff format without The MTF experimental values of the detectors
compression, using an integration time of 0.050 s for were normalized by dividing them by the same value
the scientific CCD detector and 0.004 s for the used to normalize the corresponding adjustment
CMOS detector. In the case of the video camera curve (zero-order coefficient in the polynomial-fit
CCD detector, the frames were extracted in tiff expression).
format from a video recording captured at a rate of
The results are shown in Figures 2 and 3,
25 frames/s for 1 s.
which reflect, for the different wavelengths of the
When a FFT is performed on a dataset of visible spectrum, the experimental values of the
length N, the Nyquist frequency appears at the N/2 horizontal MTF of each detector after normalization
component of the FFT output. A ratio can be formed at zero spatial frequency. For greater clarity, all the
to evaluate the spatial frequency ξn that corresponds points corresponding to the 512 MTF experimental
to the n’th component as20: values are not shown.
ξ Ny ξn
= (5)
N 2 n
Eq. (5) associates frequencies between zero and
the Nyquist frequency with FFT components from 0
to the N/2 component. In this work, we used
N=1024, thus the total number of spatial frequencies
contained in the range from 0 to the Nyquist
frequency of the detector was 512.
Before processing, each digitized frame of
speckle data was corrected in order to reduce effects
from the spatial noise of the detector itself.
With respect to the spatial noise of a CCD, a
distinction can be made between the fixed pattern Fig. 2. MTF experimental values of the scientific camera
noise (FPN) and the photoresponse non-uniformity CCD detector at different wavelengths of the visible
(PRNU). The FPN refers to the pixel-to-pixel spectrum.
variation that occurs when the array is in the dark,
and thus it is signal-independent noise. The PRNU is
due to the difference in response of each pixel to a
given signal; it is therefore signal-dependent noise.
The FPN was corrected by subtracting from the
speckle image the dark image captured obscuring the
detector, and the PRNU by means of the procedure
proposed elsewhere15.
For the processing of the speckle images, the
appropriate software was developed using
MATLAB.
4. Results and discussion Fig. 3. MTF experimental values of the scientific camera
For each wavelength analysed, the experimental CMOS detector at different wavelengths of the visible
values of the horizontal MTF of the detectors were spectrum.
calculated using Eq. (1).
For the two detectors, at each wavelength Figures 4 and 5 show, for the two detectors and
analysed, a polynomial fit of the experimental MTF at the different wavelengths, the MTF curves given
values was made and the resulting functional by the polynomial adjustments of the experimental
expression was normalized by dividing it by the values, after normalization at zero spatial frequency.
value that the adjustment equation provided at zero At the different wavelengths, MTF curves of
frequency. the scientific camera CCD detector were determined

by fitting experimental values to a second-order TABLA III

polynomial function. The correlation coefficient Squared correlation coefficients of fitting curves shown in
corresponding to each of these adjustments is Fig. 4.
indicated in Table III. Squared
Wavelength (nm) correlation
For the scientific camera CMOS detector, MTF
coefficient
curves were determined by fitting experimental
632.8 0.9632
values to a third-order polynomial function at each
514 0.9305
wavelength. The correlations coefficients associated
488 0.9442
with these polynomial fits are shown in Table IV.
457 0.9462
The comparison of the results show the
differences between the cameras analysed with
respect to the performance of the MTF of the TABLA IV
detector with wavelength, within the spectral range
Squared correlation coefficients of fitting curves shown in
studied. For both scientific detectors, the MTF Fig. 5.
spectral behaviour does not show the same trend.
Squared
Wavelength (nm) correlation
coefficient
632.8 0.9860
514 0.9868
488 0.9800
457 0.9598
Within the range of the visible spectrum

comprised between 457 and 514 nm, no significant
differences were appreciated in the MTF curves of
the scientific detectors resulting at the different
wavelengths, at least for the spatial frequencies
distant from the Nyquist frequency, as stated in
previous works24,25.
Fig. 4. MTF of the scientific camera CCD detector at The overall MTF behaviour of the scientific
different wavelengths of the visible spectrum. Curves were detectors is determined by the geometrical shape of
determined by fitting experimental values to a second- the active pixel area and the physical-diffusion
order polynomial function.
effect26,27. The diffusion component of the MTF is
due to the penetration depth of photons into the
substrates and, as the wavelength increases, photon
absorption occurs at increasing depths in the detector
material26,27. It was therefore expected that the MTF
of the scientific detectors changes more clearly at
longer wavelengths than those analysed in the works
mentioned formerly. This is why, in the present
work, we are interested in extending the study of the
MTF spectral-variation of the different detector
arrays at higher wavelengths within the visible
spectrum.
The new measurements carried out show that,
for the scientific CMOS detector, the differences
between the MTF curves corresponding to the
Fig. 5. MTF of the scientific camera CMOS detector at different wavelength augment with the spatial
different wavelengths of the visible spectrum. Curves were frequency. Specifically, the MTF decreases as the
determined by fitting experimental values to a third-order wavelength increases for the spatial frequencies
polynomial function. beyond the middle of the interval analysed (that is,

approximately from half of the Nyquist frequency of For the scientific CCD detector, the differences
the detector). between the MTF curves corresponding to extreme
wavelengths are not likely to be significant; but they
Within the whole spectral range studied here,
are for the scientific CMOS detector.
the MTF spectral behaviour of the scientific
detectors does not show the same trend, the Since the penetration depth of photons in the
differences between the MTF curves at the different detector material increases with the wavelength,
wavelengths being more notable for the CMOS than effects of smear and signal loss are more
for the CCD scientific detector. Even the differences pronounced at higher wavelengths. Therefore, the
between the MTF curves corresponding to extreme influence of the diffusion on the MTF of the detector
wavelengths are probably not significant for the is stronger as the wavelength increases.
scientific CCD, as opposed to what appears for the
The new measurements incorporated into the
CMOS detector.
present work demonstrate this fact, as the results
Differences in the MTF spectral behaviour of found for the scientific CMOS detector illustrate. In
the detector arrays studied could be due to the effect the case of the scientific CCD detector, to observe
of charge diffusion between pixels, which depends these effects, it would be necessary to carry out new
on wavelength26. Probably, the charge diffusion measurements at higher wavelengths than those
effect is slightest for the CCD scientific camera, and analysed here.
therefore the wavelength does not influence the
In this sense, it is worthwhile to become aware
MTF of its detector significantly.
of the differences reported in the MTF spectral-
On the other hand, our results show that the behaviour of the detector arrays analysed in our
CCD detector presented MTF values higher than work, taking them into account when choosing a
those of the CMOS detector at the same spatial solid-state camera for specific applications within
frequencies, for each of the four visible wavelengths the spectral range we have studied.
analysed.
Acknowledgments
5. Conclusions The authors express their appreciation to the
In this work, we have comparatively analysed the Ministerio de Educación y Ciencia for financing
quality of the images provided by different detector projects FIS2004-06465-C02-02 and FIS2007-
arrays using the speckle method. In addition, we 66671-C02-02.
have compared their performance with wavelength
within a range of the visible spectrum. For this, we
have studied the MTF at several visible
wavelengths, for the detectors of two scientific
cameras (a CCD and a CMOS).
For all the wavelengths studied, the scientific
CCD detector presented MTF values higher than
those of the CMOS detector.
Moreover, our results reveal differences in the
MTF spectral-variation of the detector arrays
analysed within the spectral range studied.
In the case of both scientific cameras, the MTF
spectral behaviour does not show the same trend
within the spectral range studied, the differences
between the MTF curves being more notable for the
CMOS detector.
For the scientific CMOS detector, the
differences between the MTF curves arise as the
spatial frequency augments. Beyond half of the
Nyquist spatial frequency of this detector, the MTF
decreases as the wavelength increases.

Color in virgin-olive-oil tasting glasses
L. Gómez-Robledo (1), M. Melgosa (1), F. J. Heredia (2),

M. J. Moyano (3), R. Roa (1), R. Huertas (1)
1. Departamento de Óptica. Facultad de Ciencias. Universidad de Granada. 18071 Granada, Spain.

2. Grupo de Color y Calidad de Alimentos. Facultad de Farmacia.
Universidad de Sevilla. 41012 Sevilla, Spain
3. Almazara Experimental del Instituto de la Grasa, CSIC. 41012 Sevilla, Spain.
Email: luisgrobledo@ugr.es
ABSTRACT:
Color measurements of 18 virgin-olive-oil tasting glasses and 10 different virgin olive oils, in a
color cabinet with a D65 source. These measures were made using a spectroradiometer with
geometries tilted at 0º, 30º, and 60º, simulating different positions of the taster’s eye. None of the
cups employed had all their geometrical dimensions within the standardized values, despite being
cups used in official sensorial analyses. Comparing color variability of the oils in different tasting
cups we discovered that in all the color differences the average color difference was above visual
threshold.
Keywords: Virgin olive oil, oil tasting, color measurement, color differences, standard oil-tasting
cups, MCDM.

[1] R. Gutiérrez, F. Gutiérrez, “Método rápido para definir el color de los aceites de oliva vírgenes”, Grasas y
Aceites, 37, 282-284 (1986).
[2] M. J. Moyano, M. Melgosa, J. Alba, E. Hita, F. J. Heredia, “Reliability of the bromthymol blue method for
color in virgin olive oils”, J. Am. Oil Chem. Soc., 76, 687-692 (1999).
[3] Commission on Illumination (CIE). CIE Publication 15 2004. Colorimetry (Technical Report). 3rd ed.
International CIE Central Bureau (2004).
[4] M. Melgosa, M. M. Perez, E. Hita, F. J. Heredia, J. Alba, M. J. Moyano, “Precision and accuracy in the
color specification of virgin olive oils from the bromthymol blue method”, J. Am. Oil Chem. Soc., 77,
1093-1099 (2000).
[5] M. Melgosa, R. Huertas, E. Hita, J. M. Roa, F. J. Heredia, J. Alba, M. J. Moyano, “Proposal of a uniform
color scale for virgin olive oils”, J. Am. Oil Chem. Soc., 81, 323-329 (2004).
[6] L. Gómez-Robledo, M. Melgosa, R. Huertas, R. Roa, M. J. Moyano, F. J. Heredia, “Virgin-olive-oil color
in relation to sample thickness and the measurement method”, J. Am. Oil Chem. Soc., 85, 1063-1071
(2008).
[7] Consejo Oleícola Internacional, COI 1987. Copa para la degustación de aceites., COI/T.20/DOC. (1987).
[8] R. Roa, R. Huertas, M. A. López-Álvarez, L. Gómez-Robledo, M. Melgosa, “A comparison between
illuminants and light-source simulators”, Opt. Pur. Apl., 41, 291-300 (2008).

1. Introduction We poured 200 ml of oil in each standard cup,

Color is not considered in virgin-olive-oil sensorial and located the cup in a fixed position on the floor of
analyses, but it is recognized as an important a GretagMacbeth Spectralight III color cabinet
organoleptic property influencing consumer’s equipped with a daylight source which simulates the
preferences and subsequent choices. Therefore CIE D65 illuminant quite well8. Spectral radiant
different color-specification techniques for virgin powers were measured using a SpectraScan PR704
olive oils have been reported in the literature. Thus, (Photoresearch Inc., Chatsworth, CA, USA)
a visual comparison between the color of a given oil spectroradiometer. The spectroradiometer was
sample and a two-dimensional scale with 60 fixed positioned in front of the color cabinet on a sturdy
solutions, called bromthymol blue (BTB) scale, has tripod that allowed both vertical movements and
been suggested1. The BTB standards were correlated three different tilts of the optical axis of the
with CIELAB color coordinates2, because currently spectroradiometer: 0º, 30º, and 60º (see Figure 1).
CIELAB is one of the two internationally proposed These different positions of the spectroradiometer
spaces for color specification3. The limited precision may be considered representative of the ones
and accuracy achievable from the BTB standards4 occupied by the taster’s eyes performing sensorial
led to the proposal of an alternative theoretical scale, analyses in a cabinet. The spectroradiometer
named Uniform Oil Color Scale (UOCS), with performed spectral power measurements in the range
improved performance upon BTB method5. Color 380 to 780 nm at 2 nm steps, using a measurement
specification of virgin olive oils from field of 1º. The CIE 1964 Supplementary Standard
spectrophotometric and spectroradiometric Observer3 was assumed in all our computation.
measurements have also been proposed, showing the
lack of good correlation between these two
techniques6. Following our previous research on
color of virgin olive oils, this paper focuses on the
color of standard virgin-olive-oil tasting cups,
analyzing whether they really conceal the color of
commercial virgin-olive-oil samples. From the
results, we provide information on the standard cups
currently employed in official virgin-olive-oil
tasting7, seeking to improve virgin-olive-oil
sensorial analyses.
2. Materials and Methods

We have collected 18 blue-tinted standard cups for
virgin olive oils, employed by experts at different
laboratories of the “Instituto de la Grasa” (National
Research Council, Seville, Spain). A vernier caliper,
with sensitivity ± 0.05 mm, was used to measure the
different dimensions of each standard cup, and a
glass-graduated cylinder, with a sensitivity of ± 2.0
ml, to measure their capacity.
Fig.1. Scheme of the setup for spectroradiometric
10 commercial extra-virgin olive oils produced measurements.
in Spain were selected, which might be considered
representative of the extra-virgin olive oils found by
consumers in the marketplace. The spectral
3. Results and discussion
transmittances of samples of these 10 oils were For each geometry (0º, 30º, and 60º) we have
measured with a JASCO-V650 spectrophotometer obtained a total of 13,770 CIELAB color
(Jasco Europe S.R.L., Cremella, Italy), using a 5.0 differences, the result of considering all the
mm path-length cell, and their corresponding combinations by pairs of the 18 available tasting
CIELAB color coordinates were computed, cups (153 pairs) and the 10 virgin olive oils (45
assuming the D65 illuminant and CIE 1964 pairs), multiplied by 2 because we must distinguish,
Supplementary Standard Observer. for example, the case of oil 1 in cup 1 versus oil 2 in

cup 2, from the case of oil 1 in cup 2 versus oil 2 in Table I shows the average CIELAB color difference
cup 1.
( ΔE )
*
ab i , j
and standard deviation (SD) found with
Figure 2 shows the percentage of color pairs
with CIELAB color differences in different the 13,770 color pairs possible from our 18 oil-
intervals, and in the different geometries of 0º, 30º, tasting cups and 10 virgin olive oils. In agreement
with previous results, Table 4 shows that the lowest
and 60º. For the 0º geometry the percentage of color
average color differences and standard deviations
pairs with CIELAB color differences in the interval
correspond to the 0º geometry, with values clearly
0-2 CIELAB units is close to 50%, while this
greater than the human visual threshold. In any case,
percentage is considerably lower (around 15%) for it must be said that, as expected, the use of blue-
the 30º and 60º geometries. A large percentage of tinted glasses in oil-tasting cups reduces the
small color differences (< 2 CIELAB units) can be perceived color differences between different olive
found for two tasting cups with different oils when oils: for example, in a previous work6 we found that
the visual observation is made in a horizontal when transparent Pyrex glasses with 46.4 mm
direction. This percentage of small color differences thickness were used, the average color differences
is lower than for 30º or 60º visualization. A very from the 10 oils considered here was 10.96 CIELAB
high number of pairs with clearly perceptible color units, with a standard deviation of 7.23 CIELAB
differences (> 2.0 CIELAB units): specifically, more units (clearly higher than those shown in Table 4).
than 50% of the total pairs for the 0 º geometry and
more than 85% of the total pairs for the 30 º and 60 º
geometries. There also are pairs with very strong TABLE I
color differences (>15 CIELAB units). Average CIELAB color difference and standard deviation
(SD) computed from the 13,770 color pairs obtainable
from our 18 olive-oil tasting cups and 10 virgin-olive oils,
considering each one of the 3 measurement geometries.
Angle ΔE SD
ab,10
50
0deg
30deg
0º 3,07 2,87
40 60deg
30º 5,25 3,07
30
60º 6,89 4,59
%
20
10 4. Conclusion
0
The instrumentally measured color-differences in
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 this work are greater than typical human visual
ΔE*ab,10 (CIELAB units)
thresholds in many cases. It would be reasonable in
virgin-olive-oil sensorial analyses to propose the use
Fig.2. Histogram of CIELAB color differences measured
of opaque tasting cups made with black material. In
by a combination of all potential color pairs from 18
tasting cups and 10 virgin olive oils, distinguishing the this way it would be possible to completely avoid
three measurement geometries (0º, 30, and 60º). the influence of color on the tasters’.
º
Acknowledgments
Research Project P06-AGR-01744, Consejería de
Innovación, Ciencia y Empresa, Junta de Andalucía
(Spain), with European Regional Development Fund
(ERDF) support.

Iris color evaluation through multispectral systems
J. Herrera, M. Vilaseca, J. Pujol, M. Arjona, M. de Lasarte
Dept. de Óptica y Optometría. Universidad Politécnica de Cataluña (UPC). Centro de Desarrollo de Sensores,
Instrumentación y Sistemas (CD6). (http://www.cd6.upc.es/).
E-mail: Jorge.alexis.herrera@cd6.upc.edu
RESUMEN:
En este trabajo se hace un estudio de las herramientas multiespectrales en la aplicación de la
medición del color en iris humanos, prótesis oculares y lentes de contacto cosméticas. Con el
análisis estadístico de los valores CIELa*b*, las diferencias de color CIEDE2000 y las gamas de
color las muestran se clasifican en tres grandes grupos: marrón, azul y verdes. Así mismo, se
analiza la reproducción que de los iris humanos hacen las prótesis y las lentes de contacto. Los
resultados muestran que los iris humanos utilizados en este estudio son predominantemente
marrones y que la reproducción del color hecha por las prótesis es aceptable mientras que no es
así en el caso de las lentes de contacto. Además, se plantea como perspectiva el incluir el análisis
de no sólo los valores colorimétricos promedio, sino también el análisis de la distribución espacial
del color.
ABSTRACT:
In this work a study of multispectral tools in color measurement of irises, prostheses and cosmetic
colored contact lenses is performed. With the statistical analysis of CIELa*b* values, the
CIEDE2000 color differences and color gamuts, the samples are classified in three major groups:
brown, blue and green. In the same way, the irises color reproduction made by prostheses and
contact lenses is analyzed. The results show that brown irises are predominant into the set of
samples utilized and the irises color reproduction in prostheses is closer than in contact lenses.
Besides, we set out as perspective, to include the analysis of not only mean colorimetric values,
but also the analysis of color spatial distribution.
Keywords: Color, multi-spectral systems, iris, gamut.

[1] W. E. Budde, I. M. Velten, and J. B. Jonas, “Optics disc size and iris color”, Arch. Ophthalmol., 116, 545
(1998).
[2] M. Seddon, C. R. Sahagian, R. J. Glynn, R. D. Sperduto, and E. S. Gragoudas, “Evaluation of an iris color
classification system. The eye disorders case-control study group”, Invest. Ophthalmol. Visual Sci., 31,
1592–1598 (1990).
[3] E. J. German, M. A. Hurst, D. Wood and J. Gilchrist, “A novel system for the objective classification of iris
colour and its correlation with response to 1% tropica-mide”, Ophthalmic Physiol. Opt., 18, 103-110
(1998).
[4] M. Melgosa, M. J. Rivas, L. G. Mez, and E. Hita, “Towards a colorimetric characterization of the human
iris”, Ophtal. Physiol. Opt., 20, 252 (2000).

[5] M. Vilaseca, R. Mercadal, J. Pujol, M. Arjona, M. de Lasarte, R. Huertas, M. Melgosa, and F. H. Imai,
“Characterization of the human iris spectral reflectance with a multispectral imaging system”, Appl. Opt.,
47, 5622-5630 (2008).
[6] J. Daughman, 1994. U.S. Patent 5,291,560. Biometric personal identification system based on iris analysis.
Issue date: March 1, 1994.
[7] P. D. Imesch, I. H. L. Wallow, and D. M. Albert, “The color of the human eye: A review of morphologic
correlates and of some conditions that affect iridial pigmentation”, Surv. Ophthalmol., 41, S117-S123
(1997).
[8] E. Pascal, E. Gooding, and L. Hannan, “Is the albino iris worse than no iris?”, Ophthalmic Physiol. Opt.,
18, 383. (1998).
[9] G. Jordan, and J. D. Mollon, “Rayleigh matches and unique green”, Vis. Res., 35, 613-620 (1995).
[10] F H. Imai, Preliminary Experiment for Spectral Reflectance Estimation of Human Iris using a Digital
Camera. Technical Report at Munsell Color Science Laboratory (2002).
(http://www.cis.rit.edu/mcsl/research/reports.php).
[11] M. Vilaseca, M. de Lasarte, J. Pujol, M. Arjona, and F. H. Imai, “Estimation Of Human Iris Spectral
Reflectance Using A Multi-Spectral Imaging System”, in Proceedings of the Third European Conference
on Colour in Graphics, Imaging and Vision (CGIV 2006), 232-236 (2006).
[12] M. Vilaseca, M. de Lasarte, J. Pujol, M. Arjona, R. Huertas, M. Melgosa and F. H. Imai, “Measuring and
Analyzing the Colour of the Iris with a Multi-Spectral Imaging System”, in Proceedings of the Fourth
European Conference on Colour in Graphics, Imaging and Vision (CGIV 2008), 427-431 (2008).
[13] M. de Lasarte, M. Vilaseca, J. Pujol, and M. Arjona, “Color measurements with colorimetric and
multispectral imaging systems”, Proc. SPIE 6062, pg. 0F1 (2006).
[14] M. Vilaseca, J. Pujol, and M. Arjona, “Spectral-reflectance reconstruction in the near-infrared region by use
of conventional charge-coupled-device camera measurements”, Appl. Opt., 42, 1788 (2003).
[15] M. Vilaseca, J. Pujol, and M. Arjona, “Illuminant influence on the reconstruction of near-infrared spectra”,
J. Imaging Sci. Technol., 48, 111 (2004).
[16] M. J. A. Port, Cosmetic and prosthetic lenses. In: Contact Lenses, 3rd ed. Eds A. J. Phillips and J. Stone,
Butterworth-Heineman Ltd, pp. 789-796 (1989).
[17] M. de Lasarte, J. Pujol, M. Arjona, and M. Vilaseca, “Optimized algorithm for the spatial non-uniformity
correction of an imaging system based on a CCD color camera”, Appl. Opt., 46, 167–174 (2007).
1. Introduction cosmetic colored lenses in three different groups:

Iris color has been studied in just some works . 1-5 brown, blue and green. This algorithm evaluates
Some of these studies have been subjective some criteria over the CIELa*b* coordinates of each
observations1,2 and few have attempted quantitative sample. These coordinates are obtained by means of
measurements using standard color instrumentation3- an optimized and calibrated multispectral system
5
. This shows that quantitative iris color product of former works5,10-15.
measurement is a task neither ended nor easy. Statistical analyses, color differences and
Although irises have such variability in color and gamut comparisons are performed as a first approach
texture that make them suitable for irises to the color characterization of samples as well as
identification applications6 and implicitly set exact the comparison of prostheses and contact lenses
color measurements as almost impossible ones, accounting their color characteristics in respect to
increasing the accuracy in quantitative color and irises. In that way, we attempt to obtain objective
spectral reflectance characterizations are desirable color data that could assist prostheses and contact
for other several applications7-9. Thus, some recent lenses producers in their goal of reproducing iris.
works have been implementing new colorimetric Task that nowadays is performed by trial and error
elements like color image processing and with subjective comparisons16.
multispectral systems to face these studies5,10-12.
Following this tendency this work classifies, through With this work we show the different gamuts
an automatic algorithm, a set of samples composed of the three major groups resulting from the
by 106 human irises, 68 ocular prostheses and 17 proposed classification. It clearly shows the group of
brown color as the most populated for the irises and

prostheses samples. In the same way, color gamuts 2.b. Methods

and color differences show the color proximity
With the described system, we measured the
between irises and prostheses and also the
mean digital output levels of the samples, i.e. irises,
differences with the contact lenses. It reveals that
prostheses and contact lenses, corresponding to two
contact lenses are clearly influenced by the eye
square areas of approximately 1×1 cm on the iris
utilized in the measurement procedure, so is needed
(Fig. 2), which presented a rather uniform
a different treatment for its complete
coloration. At the same time, the corresponding
characterization.
averaged spectral reflectances in the visible range
This paper is structured as follows: in the (380 – 780 nm) of those zones were also measured
second section a brief description of the system, using a spectrophotometer (Photo Research PR-
methods and statistical tools used are presented. The 650).
results and comparison are shown in the third
section and, in the last section, the conclusions of
this work are presented.
2. Experimental setup and methods

2.a. Experimental setup Fig. 2. Image of an iris, a prosthesis and a coloured contact
lens, with the corresponding 2 analyzed areas.
The system utilized in this work is the result of
former studies that have yielded an optimized
experimental setup with its necessary computational Thus, from these digital levels we obtained the
tools5,11-15,17. This multispectral system (Fig. 1) colorimetric L*a*b* or XYZ tristimulus values
consists of a 12 bit depth cooled CCD monochrome following some former steps. Basically these steps
camera (QImaging QICAM Fast1394) with 1.4 involve:
megapixels (1392 × 1040), an objective zoom lens • System optimization: the selection of the
(Nikon AF Nikkor 28–105 mm), and a color RGB best set of filters. For our system an RGB
tunable filter attached to the CCD camera. tunable filter had been determined as the
best solution.
• Image preprocessing: flat-field correction
applied in order to both correct the camera
response and the nonuniformity of the
illumination.
• Training set calibration: tests with different
training sets for spectral profiles
reconstruction. A subset of real irises has
been used as the final training set.
• Reconstruction of spectral reflectances
profiles: tests of different reconstruction
algorithms. Moore-Penrose pseudoinverse
algorithm had been selected because of its
accuracy and implementation simplicity.
• With those spectral reflectances profiles
Fig.1. Multispectral experimental setup.
and the illuminant D65 the XYZ and
CIELa*b* values are obtained.
The reasons and details of the former
Furthermore, there is an illumination system procedures are deeply explained in REFs5,11-13,17.
composed of an adjustable halogen lamp (Philips
15V 150 W) attached to a stabilized dc power supply
(Hewlett Packard 6642A) and a focusing lens,
allowing illumination of the analyzed iris with a 45°
angle of incidence and providing a rather uniform
luminous field on the eye.

TABLE I colored contact lenses, helping us to conclude how

Statistical CIELa*b* values for the irises samples. they replicate real irises.
a* b* L* Fig. 4 has an example showing how the
Irises Irises Irises volumes of each sample belonging to the group of
Mean 6,33 14,97 27,57 blue color overlap. It is clear how for these blue
Std. Dev. 5,77 8,77 8,65 subsets the prostheses and irises overlap much more
between them than with the contact lenses. This
Min -6,43 -5,47 10,88
behaviour gets repeated when this comparison is
Max 23,65 29,76 47,64 performed again in the other two groups of colors.
Volume 6670
The automatic classification of the samples is

performed based on statistical descriptors extracted
from their CIELa*b* values. The table I shows the
main descriptors for the group of samples coming
from real irises. With it we can identify the ranges in
each colorimetric coordinates, their dispersions and
besides, we can analyze volumes, namely, color
gamuts.
3. Results and comparison

With the analysis we built boundaries in the
CIELa*b* space for the separation of the samples Fig. 4. Comparison of blue gamuts.
which give us the following gamuts when they are
applied to the irises samples (Fig. 3).
This is a first approach that shows how prostheses
and irises are more similar between them. The same
conclusion is evident when we use the data extracted
from color differences, the mean values and the
dispersion of the minimum color differences
between contact lenses samples and irises are bigger
than between prostheses and irises (TABLE II). It
shows that prostheses are a better representation of
the irises than the contact lenses regarding to their
mean color values. It can be said that taking into
account how the gamuts of each color group are
represented by the prostheses or contact lenses, the
prostheses cover almost completely the gamuts of
the irises for each color group, it means that is very
likely to find some close representation of a desired
iris in the set of prostheses, at least regarding to their
Fig. 3. CIELa*b* volumes for the proposed classification bulk color appearance. But, on the other hand, it sets
in real irises. the need of more accurate color analysis, not only
including mean colorimetric values for a determined
region, but including a color distribution analysis
As can be seen, this separation shows the brown
taking advantage of the obtained data over the whole
group as the biggest and most populated one. Similar
image through the multiespectral system.
results are obtained in prostheses and contact lenses.
Using this classification we can compare the
samples by colors and analyze the color
reproduction in prostheses and colored contact
lenses with respect to real irises. In addition,
utilizing the CIEDE2000 formula for calculating the
color differences we found the nearest pair for each
iris in the other two sets of samples, prostheses and

TABLE II
CIEDE2000 minimum color differences with respect to
real irises. Irises-Prostheses(I-P), Irises-Lenses(I-L).
∆E00 I-P ∆E00 I-L
Mean 2.28 4.47

Std.
Dev 1.02 3.95
4. Conclusion and perspectives

This work shows the results of an automatic
algorithm to classify a set of samples composed by
106 real irises, 68 prostheses and 17 colored contact
lenses in color groups: brown, blue and green. This
classification shows the brown group as the most
populated one and with the biggest volume in the
CIELa*b* color space. This classification allows us
a deeper analysis of the irises color reproduction
through the representation of gamuts in each color
group and the observation of their overlapping. The
overlapping shows us how the prostheses are very
close to the irises regarding their gamuts. Moreover,
with the analyses of the color differences utilizing
the CIEDE2000 formula, the same behaviour is
observed. Thus, we can state that the prostheses
make a good representation of the irises colors at
least concerning to their mean color values. The next
step will be to include the analysis of spatial color
distribution to classify and compare the samples
taking advantage of the multi-spectral data. In
addition, a close study of the influence of the eye in
the contact lenses measurement is intended to be
done.

Reproducción espectral de valores triestímulo mediante

descripciones B-Spline: evaluación del error en el color
C. Pizarro, J. Arasa, M. de Lasarte, J. Pujol, M. Arjona, M. Vilaseca
Departamento de Óptica y Optometría. Universidad Politécnica de Cataluña (UPC). Centro de Desarrollo de

Sensores, Instrumentación y Sistemas (CD6). Rambla Sant Nebridi 10, 08222 Terrassa (Barcelona - España).
Email: marta.lasarte@oo.upc.edu
RESUMEN:
La principal motivación de este trabajo es la búsqueda de una única expresión matemática que
permita reproducir distribuciones espectrales de forma general. Para ello se consideran
polinomios B-Spline rotacionales de segundo orden como expresión matemática base para dicha
reproducción. El objetivo fundamental de este trabajo es, por tanto, la determinación de los
coeficientes de los polinomios B-Spline que permitan reproducir distribuciones espectrales, así
como la evaluación de la exactitud de la reproducción espectral. El método utilizado para
determinar estos coeficientes es un proceso de optimización genético basado en el Algoritmo
Simplex, usando el error cuadrático medio o RMSE como función de mérito. La calidad final de
la reproducción se evalúa mediante el ‘Goodness of Fit Coefficient’ o GFC.
Palabras clave: Reproducción espectral, valores triestímulo, polinomios B-Spline.

[1] M. Ares, S. Royo, J. Caum, C. Pizarro, “Comparison of B-Spline and Zernike fitting techniques in complex
wavefront surfaces”, in Proceedings of SPIE ISBN: 0-8194-5856-2 (2005).
[2] C. Pizarro, J. Arasa, N. Tomàs, “New optimization process based on the contrast function to triplet systems
design”, in Proceedings of SPIE ISBN: 0-8194-5133-9 (2004).
[3] J. Arasa, C. Pizarro, “Aplicación de algoritmos genéticos para obtener la mezcla óptima de colores básicos
en los procesos de reproducción del color”, in Proceedings del VII Congreso Nacional de Color, Pamplona
19 – 21 Mayo (2004).
[4] J. Arasa, C. Pizarro, “Use of aplanatic condition to sample extended objects: application to merit function
segmentation”,in Proceedings of SPIE ISBN: 0-8194-5980-1 (2005).
1. Introducción la determinación de los coeficientes de los

La principal motivación de este trabajo es la polinomios B-Spline que permitan reproducir
búsqueda de una única expresión matemática que distribuciones espectrales, así como la evaluación de
permita reproducir distribuciones espectrales de la exactitud de la reproducción espectral. En trabajos
forma general. Para ello se consideran polinomios anteriores se ha llevado a cabo la reproducción de la
B-Spline rotacionales de segundo orden como distribución espectral de algunos iluminantes
expresión matemática base para dicha reproducción. comprobando la validez del método. En este trabajo
El objetivo fundamental de este trabajo es, por tanto, se presentan los resultados obtenidos para la

reproducción espectral de los valores triestímulo de representan el método geométrico de determinación

un conjunto de muestras de colores, evaluando la de los coeficientes a partir de las curvas base, y la
exactitud de la reproducción mediante la diferencia línea gris que une los puntos de control define el
de color CIELAB entre los valores triestímulo lugar geométrico del espacio donde se sitúa el
reproducidos y los medidos directamente sobre las polinomio B-Spline. Se puede encontrar una
muestras. descripción matemática completa en la referencia1.
En la Figura 2 se puede observar un ejemplo de
aplicación de los polinomios B-Spline en la
2. Método experimental reproducción de distribuciones espectrales.
En este trabajo se consideran polinomios B-Spline
rotacionales de segundo orden como expresión
matemática base para la reproducción de
distribuciones espectrales. Estos polinomios se
caracterizan por ser continuos y derivables en los
extremos, cosa que permite enlazar polinomios con
continuidad en las segundas derivadas en sus
extremos, haciendo posible así la descripción de
curvas por tramos.
La expresión típica de un polinomio B-Spline
de orden k-1 consiste en una combinación lineal de
unas funciones base Bi,k(λ), mediante unos
coeficientes escalares ai,k según:
n
Pk −1 (λ ) = ∑ ai ,k Bi ,k (λ ) (1)
i =o
(a)
Las funciones base Bi,k(λ) presentan la forma:
λ − ti t −λ
Bi ,k (λ ) = Bi ,k −1 (λ ) + i + k Bi +1,k −1 (λ )
ti + k −1 − ti ti + k − ti +1 (2)
y se caracterizan por estar descritas por un
parámetro t que recorre la curva a reproducir.
La ventaja principal de los polinomios B-
Spline es que admiten una representación
geométrica de sus coeficientes escalares, también
denominados puntos de control (Fig. 1.).
(b)
Fig. 2. Ejemplo de reproducción de una distribución
espectral (azul) mediante un polinomio B-Spline de orden
4 (rojo) y sus puntos de control (negro). El desplazamiento
de un punto de control (a) modifica la curva sólo alrededor
del punto (b).
Fig. 1. Representación geométrica de un polinomio B-
Spline de orden 4.
En este ejemplo aparecen representadas en azul
la distribución espectral a reproducir, en rojo la
En el caso de la representación geométrica de
curva B-Spline y en negro sus coeficientes (5,
la Fig.1, la curva roja corresponde a la curva B-
polinomio de orden 4) o puntos de control. La línea
Spline descrita por un polinomio de orden 4. Los 5
negra que los une describe el lugar geométrico
puntos P0, P1, P2, P3 y P4, son la representación
donde se sitúa la curva B-Spline.
gráfica de los coeficientes o puntos de control del
polinomio B-Spline, las líneas verde, azul y rosa

El objetivo es determinar los coeficientes que En cuanto al iluminante A, en la Figura 4.(a) se

permiten reproducir la distribución espectral original representan la curva de distribución espectral
mediante una curva B-Spline, desplazando los original (en negro) y reproducida (en rojo). Tal y
puntos de control y/o incorporando nuevos puntos de como se observa, se obtiene una reproducción
control intermedios hasta obtener el ajuste deseado. excelente (RMSE = 0.57182, GFC = 0.99999)
utilizando un polinomio B-Spline de orden 4, con
Tal y como se puede observar, el
sólo 5 coeficientes o puntos de control.
desplazamiento de un punto de control (Fig. 2.(a))
sólo afecta a la curva B-Spline en la zona cercana al
punto (Fig. 2.(b)), con una influencia casi nula en el
resto de zonas alrededor del resto de puntos de
Objetivo Resultado
control no adyacentes. alcanzado
Los coeficientes del polinomio B-Spline que

permite reproducir una cierta curva se determinan a
partir de un algoritmo genético basado en el
Algoritmo Simplex. El Algoritmo Simplex también
tiene una representación geométrica en el espacio, al
igual que los coeficientes de los polinomios B-
Spline, y en él se utilizan los 4 movimientos que
aparecen representados en la Figura 3: reflexión,
reflexión y extensión, contracción y múltiple
contracción.
RMSE = 0.57182, GFC = 0.99999
Fig. 4. Distribución espectral original (negro) y
reproducida (rojo) para el iluminante A.
reflexión reflexión y extensión
En la Figura 5.(a) se pueden observar lo

resultados para el iluminante D65 utilizando un
polinomio B-Spline de orden 4 con 5 coeficientes.
contracción múltiple contracción
La reproducción no es muy buena utilizando 5
coeficientes y un polinomio de orden 4, aunque los
Fig. 3. Representación geométrica de los 4 movimientos resultados mejoran con 10 coeficientes y un
utilizados en el Algoritmo Simples. polinomio de orden 9 (Fig. 5. (b)). Con 15
coeficientes (Fig. 5.(c)) continúa la mejora aunque la
reproducción todavía no es excelente (GFC =
La descripción completa del proceso de
0.99965). Finalmente, con 30 coeficientes (Fig.
optimización genético utilizado se puede encontrar
5.(d)) se obtiene una reproducción excelente de la
en las referencias2,3,4.
distribución espectral del iluminante D65.
La función de mérito utilizada en el proceso de
optimización genético es el error cuadrático medio o
RMSE. La calidad final de la reproducción se evalúa
mediante el ‘Goodness of Fit Coefficient’ o GFC,
Objetivo Resultado
que es un parámetro notablemente diferente al alcanzado
utilizado como función de mérito, con lo que se
pretende obtener un contrapunto en la evaluación de
la calidad de la reproducción.
3. Resultados
3.a. Antecedentes: reproducción espectral de
iluminantes
A continuación se exponen brevemente los
antecedentes a este trabajo, consistentes en la
reproducción espectral de los iluminantes A, D65, Fig. 5.(a): RMSE = 5.13338, GFC = 0.99835
F2, F7 y F11.

En la siguiente tabla (TABLA I) se presentan

los resultados obtenidos en términos del RMSE y del
GFC en función del número de coeficientes
utilizados para los iluminantes A, D65, F2, F7 y
F11.
Se observa que para el iluminante A, con una
distribución espectral suave y uniformemente
creciente, son suficientes 5 coeficientes para obtener
una reproducción excelente, mientras que para los
iluminantes D65, F2, F7 y F11, con distribuciones
espectrales abruptas con numerosos picos, son
necesarios 30 coeficientes para obtener una
reproducción excelente, o bien más de 15 para una
Fig. 5.(b): RMSE = 3.43171, GFC = 0.99916 reproducción aceptable. Todo depende del grado de
precisión que se requiera en la representación.
TABLA I
RMSE y GFC para la reproducción de las distribuciones
espectrales de los iluminantes A, D65, F2, F7 y F11 en
función del número de coeficientes del polinomio B-
Spline utilizado.
Iluminante # coefs. RMSE GFC
5 0.57182 0.99999
10 -- --
A
15 -- --
30 -- --
5 5.13338 0.99835
10 3.43171 0.99916
Fig. 5.(c): RMSE = 2.36153, GFC = 0.99965 D65
15 2.36153 0.99965
30 0.85993 0.99995
5 9.21131 0.97845
10 6.34761 0.99713
F2
15 2.61352 0.99912
30 0.95830 0.99992
5 11.4456 0.88137
10 8.42169 0.99538
F7
15 3.01522 0.99901
30 1.01877 0.99993
5 6.44981 0.91536
10 5.17431 0.99715
F11
15 2.97636 0.99911
30 0.89593 0.99991
Fig. 5.(d): RMSE = 0.85993, GFC = 0.99995

3.b. Reproducción espectral de valores
Fig. 5. Distribución espectral original (negro) y
reproducida (rojo) para el iluminante D65 triestímulo
utilizando (a) un polinomio B-Spline de orden 4 A continuación se presenta un ejemplo de
con 5 coeficientes, (b) un polinomio B-Spline de
aplicación de los polinomios B-Spline en la
orden 9 con 10 coeficientes, (c) un polinomio B-
Spline de orden 14 con 15 coeficientes y (d) un reproducción espectral de valores triestímulo
polinomio B-Spline de orden 29 con 30 utilizando las 24 muestras de color de la carta
coeficientes GretagMacbeth ColorChecker Color Rendition
(CCCR).
Disponiendo de las distribuciones de
reflectancia espectral de las muestras de la carta
CCCR medidas en condiciones de iluminación

uniforme, se lleva a cabo su reproducción mediante TABLA II

polinomios B-Spline. Para cada muestra se Reproducción mediante polinomios B-Spline de las 24
determinan los coeficientes B-Spline necesarios para muestras de la carta CCCR: número de puntos de control
necesarios, valores triestímulo XYZ obtenidos a partir de
la reproducción espectral de su reflectancia. A partir
las reproducciones, y diferencias de color CIELAB entre
de la reflectancia espectral reconstruida, se calculan los valores triestímulo medidos y los obtenidos a partir de
los valores triestímulo asociados a cada muestra. la reproducción espectral con polinomios B-Spline
En la Figura 6 se representan los valores
triestímulo obtenidos a partir de la reproducción de
la reflectancia espectral de cada una de las muestras
de la carta CCCR.
4. Conclusiones
Se han aplicado los polinomios B-Spline en la
Fig. 6. Representación de los valores triestímulo de las 24 reproducción espectral de los valores triestímulo de
muestras de la carta CCCR obtenidos a partir de la las muestras de la carta CCCR encontrando para
reproducción espectral de las reflectancias de cada
muestras mediante polinomios B-Spline. cada una de ellas el número de coeficientes
necesario para obtener una GFC > 0.9999.
En la siguiente tabla (TABLA II) aparecen, para A partir de estos resultados se propone como
cada una de las 24 muestras de la carta CCCR, el trabajo futuro la incorporación de un algoritmo para
número de puntos de control necesarios para la variar de forma automática el número de
reproducción de su reflectancia espectral, los valores coeficientes del polinomio B-Spline, así como
triestímulo XYZ obtenidos a partir de estas también la incorporación de nuevos criterios para
reproducciones, y las diferencias de color CIELAB evaluar la calidad de la reproducción.
entre los valores triestímulo medidos y los obtenidos
a partir de la reproducción espectral con polinomios
B-Spline.

Colorimetric properties of thermoplastic polymers coloured with

nanopigments and conventional pigments
V. Marchante (1), F. Martínez-Verdú (1), A. Marcilla (2), M. Beltrán (2)
1. Grupo de Visión y Color. Universidad de Alicante. Ctra. San Vicente s/n, 03690 Alicante.
2. Grupo de Pirólisis y Combustión de Polímeros. Universidad de Alicante.
Ctra. San Vicente s/n, 03690 Alicante.
Email: veronica.marchante@ua.es
ABSTRACT:
Nanopigments are a new type of pigments. They are hybrid materials consisting of organic dyes
and layered silicate nanoparticles1. Nanopigments are already applied to make polymeric coatings
and they had shown improvement in mechanical, thermal and stability properties of the substrate
and dyes1,2.
In this work, nanopigments had been applied in a thermoplastic polymer to obtain a coloured
nanocomposite. Nanocomposite’s physical-chemical and colorimetric properties had been
assessed. However, only colorimetric characterization is going to be presented.
Keywords: Colorimetry, nanopigments, coloured polymers.

[1] L. F. Batenburg, H. R. Fischer, Planocolors® - a combination of organic dyes and layered silicates with
nanometer dimensions. E-polymers http://www.e-polymers.org (2001).
[2] Q. H. Zeng et al., “Clay-Based Polymer Nanocomposites: Research and Commercial Development”,
Journal of Nanoscience and Nanotechnology, 5, 1574–1592 (2005).
[3] H. G. Völz, Industrial Color Testing: Fundamentals and Techniques. Wiley-VCH Verlag GmbH, (2001).
1. Introduction
Nanopigments or Planocolors® are a new kind of N
S
C H3
N
CH
C H3
3
N
S
CH3
N
C H3
CH
3
N
S C H3
N
3
CH
CH3
3
N
S
C H3
N
3
C H3
C H3
N
S
C H3
N
3
C H3
CH3
N
S H
C
3
3
N
C
N CH
N CH N C H3 N CH N CH N CH
pigmento developed by researchers of TNO-TPD

N N
N 3 N 3 N 3 N
C H3 CH3 3 C H3
CH CH CH CH
3 S
C H3 S C H3 S S C H3 S C H3S C H3
3 CH 3 CH3 CH3
CH C H3 CH3 CH3 CH
N N N N N N N N
N N N N N
3 3
3 3 3 3 3
CH CH
CH CH CH CH CH
Eindhoven (Netherland)1.
Nanopigments are hybrid materials obtained
through the combinantion of organic dye molecules Fig.1: Schematical representation of clay sheet, dye
molecule (methylene blue) and blue nanopigment.
and layered clay nanoparticles. Nanopigments gather
advantages of dyes and pigments while avoiding
their drawbacks1.

2. Nanopigments’ laboratory synthesis In order to be able to make a comparison, four

Nanopigments are obtained in a process with two different types of pigment had been used:
main steps: - An organic pigment: blue phthalocyanine (FTA)
Firstly the nano-clays have to be dispersed in a - An inorganic pigment: blue ultramarine (AU)
solvent, better in desionized water. Therefore the
nano-clay swells. This increases clay sheets’ - Two nanopigments: one synthesised in the TNO
distance and, as a consequence, reduces clay (NA-TNO) and the other in the University of
counterions’ ionic bond force. Alicante (NA-UA).
After that, a dye solution is added to the 3.b.- Procedure

dispersion. Then the ionic exchange takes place: Polymer and pigment were processed as
Dye molecules replace clay’s counterions. This follows: firstly polymer and pigment, both in
gives colour to clay nanoparticles and makes them powder state, were mechanical pre-mixed. Then the
compatible with the polymer. mixture was introduced in a kneader under these
At this moment nanopigment is formed but conditions: 180 ºC and 40 rpm, during 10 minutes.
dispersed in the deionized water. In order to isolate Finally, a thin film (approximately 1 mm of
and purify the nanopigment we have to: thickness) was made in a hot plates press.
- Wash and filter the dispersion.

- And dry the cake. It can be done by freeze 4. Measurements
drying, spray drying or in a muffle. Spectral reflectances of the black and white
substrates (ρ0w, ρ0B) and spectral reflectances of
Synthesis process is represented in Figure 2.
samples over black and white substrates (ρw, ρB) had
been measured. To measure, a spectrophotometer
Nanoclay
KONICA MINOLTA CM-26OOd (integrated
H2O deionized
sphere) had been used.
Sieving
Dispersion
+ 5. Colorimetric Parameters
Colorant solution 5.a.- Spectral Reflectances
In figure 3 it is showed spectral reflectances of
Ionic Exchange pigmented polymer samples over white substrate. In
all the samples there is a peak in the interval
between 420-520 nm because all the pigments are
Washing and Filtering
blue.
Spectral Reflectance
Drying 1,0 White Substrate
0,8
Fig. 2: Scheme of nanopigments’ synthesis at laboratory. LLFTA
LLAU
0,6
LLNA (TNO)
ρW
3. Samples’ preparation 0,4

LLNA (UA)
WHITE SUBSTRATE
Some objectives of the research are to assess the 0,2
suitability of nanopigments as pigments for
polymers and compare them with conventional 0,0
400 450 500 550 600 650 700
polymer pigments. Therefore, once nanopigments λ (nm)
were obtained, we characterized them. To do this,
we prepared coloured samples with a thermoplastic Fig. 3: Spectral reflectances of samples over white
polymer adding a mass percentage of 0.1 %w of substrate (ρWi).
pigment.
3.a. Material
Low linear-density polyethylene (LLDPE) is
the thermoplastic polymer selected because it is one
of the most widely used.

5.b.- Absorption and Scattering Coefficients 5.c.- Colour performance

From spectral reflectances, absorption and scattering To assess nanopigments’ colour performance
coefficients can be obtained applying these and compare different types of pigments, some
equations3: parameters (like hiding power, transparency and
coloring power) were calculated3. Figures 6-8 show
a=
(1+ ρ ρ )(· ρ − ρ ) + (1+ ρ ρ )(· ρ
* *
W 0,W
*
B
*
0, B
* *
B 0, B
*
0,W − ρW* ) (1) these parameters for each type of pigment.
2·(ρ ρ − ρ ρ )
* *
B 0,W
* *
W 0, B
b = (a 2 − 1) (2)
1,8 1,62
1,6
Hiding Power
1 1 − a·(ρ + ρ ) − ρ ·ρ
* * * *
(3) D (m2/L)
S= , i =W,B
i 0 ,i i 0 ,i 1,4
arc coth
b·h b·(ρ i* − ρ 0*,i ) 1,2
1,0
0,8
K = S ·(a − 1) (4) 0,6
0,4 0,23 0,24
Figures 4 and 5 show absorption and scattering 0,2
0,16
coefficients of the four pigments used for colouring 0,0

the polymer. LLFTA LLAU LLNA LLNA
(TNO) (UA)
Absorption Coefficent (K)
Fig. 6: Hiding power of coloured polymer samples for
3,5 Blue Phthalocyanine each type of pigment.
3 Blue Ultramarine
Blue Nanopigment (TNO)
2,5
Blue Nanopigment (UA)
2 9,76
K
10 Transparency
1,5 9
8
T (mL/m2)
1
7 6,42
0,5 6
5 4,43
0
4
400 500 600 700 3 2,49
λ (nm) 2
Fig. 4: Absorption coefficients of pigmented polymer 1
samples. 0
LLFTA LLAU LLNA LLNA (UA)
(TNO)
Scattering Coefficient (S) Fig. 7: Transparency of coloured polymer samples for
3,5 each type of pigment.
Blue Phthalocyanine
3
Blue Ultramarine
2,5
Blue Nanopigment (TNO)
2 1,0
S
1,5 Blue Nanopigment (UA) 0,9 Coloring Power

0,8 Φ (m2/L)
1 0,7
0,5 0,6 0,51
0,5
0 0,38
0,4 0,30
400 500 600 700 0,3
0,17
λ (nm) 0,2
Fig. 5: Scattering coefficients of 0,1
pigmented polymer samples. 0,0
LLFTA LLAU LLNA LLNA
(TNO) (UA)
In all the samples absorption contribution is Fig. 8: Coloring power of coloured polymer samples for
more important than scattering. each type of pigment.

6. Conclusion
Always considering a colorimetric point of view,
from this research main conclusions we can extract
are:
- Nanopigments’ colour performance is similar or
even better than conventional pigments.
- They could be classified in a medium situation
between organic and inorganic pigments.
- They have more transparency and colouring
power than other pigments, what can be an
advantage depending on product’s final
application.
Furthermore, they can be considered an alternative
to toxic inorganic pigments because they have better
colour performance than inorganic pigments and, as
they don’t have heavy metals in their composition,
they are not contaminant
Acknowledgements
Thanks to TNO-TPD Eindhoven for allowing us to
reproduce the synthesis of Planocolors® and for
their technical and knowledge support.
Veronica Marchante would like to thank the
“Conserlleria d’Empresa, Universitat i Ciència” for
the PhD grant that she has received.

The use of optimal stimuli in colour appearance modelling
E. Perales (1), C. Li (2), E. Chorro (1), V. Viqueira (1), F. Martínez-Verdú (1), M. R. Luo (2)
1. Departamento de Óptica, Farmacología y Anatomía. Universidad de Alicante, 03690 Alicante.

2. Department of Colour Science. University of Leeds. Leeds LS2 9JT, UK.
Email: esther.perales@ua.es
ABSTRACT:
A mathematical approach is proposed for searching the chromatic adaptation matrix for the
CAT02. This approach combines the non-negativity constraint for the corresponding colours’
tristimulus values with the minimisation of the colour differences between the tristimulus values
of the corresponding colours obtained by visual observations and tristimulus values of the
corresponding colours predicted by the model, which resulted in a constrained non-linear
optimisation problem. A new matrix has been found using the MATLAB routine “fmincon”.
The performances of the CAT02 with various matrices including the original CAT02 matrix and
the new matrix are tested using the visual data sets and the optimal colours. Test results show that
the CAT02 with the new matrix is the only one combination which successfully predicts
corresponding colours for all optimal colours. Althought we have to scarify the accuracy in order
to satisfy the non-negativity constraint for the corresponding colours.
Keywords: Chromatic adaptation, colour appearance models, optimal colours.

[1] F. M. Verdú, E. Perales, E. Chorro, D. de Fez, V. Viqueira, and E. Gilabert, “Computation and
visualization of the MacAdam limits for any lightness, hue angle and light source”, J. Opt. Soc. Am. A, 24,
6, 1501-1515 (2008).
[2] CIE 159:2004, Colour Appearance Model for Colour Management Systems: CIECAM02. CIE, Vienna,
(2004).
[3] M. R. Luo and C. J. Li, CIE Colour Appearance Models and Associated Colour Spaces, Chapter 11,
Colorimetry (understanding the CIE System). John Willey & Sons, Hoboken, New Jersey (2007).
[4] M. H. Bill, “Irregularity in CIECAM02 and Its Avoidance”, Color Res. Appl., 31 (2), 142-145 (2006).
[5] S. Süsstrunk and M. H. Bill, “The Nesting Instinct: Repairing Non Nested Gamuts in CIECAM02”, in
Fourteenth Color Imaging Conference (2006).
[6] M. H. Brill and S. Süsstrunk, “Repairing Gamut Problems in CIECAM02: A Progress Report”, submitted
to Color Res. Appl. (2007).
[7] I. Tastl, M. Bhachech, N. Moroney, and J. Holm, “ICC Color Management and CIECAM02”, in Thirteenth
Color Imaging Conference (2005).
[8] C. J Li., M. R. Luo, B. Rigg and R.W.G. Hunt, “CMC 2000 Chromatic Adaptation Transform:
CMCCAT2000”, Color Res Appl; 27:49–58 (2002).

[9] R. W. G. Hunt, C. J. Li, L. Y. Juan, and M. R. Luo, “Further Improvements to CIECAM97s”, Color Res
App., 27:164–170 (2002).
[10] CIE 15:2004, Colourimetry, 3rd ed.. CIE, Vienna (2004).
1. Introduction seen that all colours are on the CIE chromaticity

In a previous work we have developed a method to locus.
compute and visualise the gamut boundary of the
optimal colours under any light source and CIE 0.9
standard colorimetric observer1. In order to compare
the gamut boundaries under different illuminants, 0.8
the CAT022,3 was employed to transform all

0.7
boundaries under a single reference illuminant. It
was found that many of the colours from the 0.6
boundaries failed the CAT02. In other words, for
many colours from the boundaries, the CAT02 0.5
y
predicts their corresponding colours with negative 0.4
s2
tristimulus values.
0.3
On the other hand, other authors have reported
s1
problem with the chromatic adaptation transform 0.2
built in the CIECAM024,5,6. During the CIE meeting
0.1
in China in 2007, a technical committee was formed
to improve the CIECAM02. And it was found that 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
all colours which fail the CIECAM02 occur on or x
near the CIE chromaticity locus. Besides, it was
widely thought that the main problem comes from Fig.1. Chromaticity coordinates (circles) for the colours
which failed the CAT02. The full curves are the CIE
the chromatic adaptation transform, because it chromaticity locuses.
predicts the corresponding colours with negative
tristimulus values. Hence, modifications towards the
CAT02 were carried out. One of the suggestions by Therefore, in this work, we propose a general
Brill and Süsstrunk is to replace the CAT02 matrix: method for generating matrix for the chromatic
adaptation transform, which will be discussed in the
⎛ 0.7328 0.4296 − 0.1624 ⎞ next section.
⎜ ⎟
M 02 = ⎜ − 0.7036 1.6975 0.0061 ⎟ (1)
⎜ 0.003 0.0136 0.9834 ⎟⎠
⎝
by 2. Method
To modify the chromatic adaptation transform, it is
⎛ 0.7328 0.4296 − 0.1624 ⎞ necessary to know the mathematical expression for
⎜ ⎟
M BS = ⎜ − 0.7036 1.6975 0.0061⎟ (2) the chromatic adaptation transform. CAT022,3 can be
⎜ 0. ⎟
⎝ 0. 1 ⎠ compactly expressed by
Note that first two rows of the two matrices are s c = M −1ΛMs (3)
the same. The different occurs at the last row. It can
be found that for some colours the CAT02 with the where:
original matrix (eq. 1) predict corresponding colours
- s is the tristimulus values of the object colour
with negative tristimulus values, but the CAT02
under the test illuminant,
with the BS matrix (eq. 2) works fine.
- sc is the corresponding colour under the reference
The CAT02 with the MBS was also used to
illuminant.
predict the corresponding colours for optimal
colours and it also failed for many of them. Some - M is the CAT02 matrix
colours (in terms of chromaticity coordinates shown
- Λ is a diagonal matrix depending on the test and
by circles) which failed the CAT02 under different
reference illuminants and the adaptation factor and
illuminants were shown in Figure 1, where the full
is given by this expression:
curves are the CIE chromaticity locuses. It can be

fit to the experimental data, CIELAB colour

⎛ R ⎞ difference10 is used.
⎜ Dβ wr + 1 − D ⎟
⎜ Rw ⎟
⎜ ⎟ Thus, the above discussions lead to a constraint
Gwr
Λ ( D) = ⎜ Dβ +1− D ⎟ and non-linear optimization problem. MATLAB
⎜ G w ⎟
⎜ Bwr ⎟ routine “fmincon” was used for solving the problem.
⎜ Dβ +1− D⎟ 82 illuminants were used for the optimisation. The
⎝ B w ⎠
chromaticity coordinates for all illuminants are
with
shown in Figure 2. The dotted curves in Figure 2 are
Yw the CIE 1931 (black) and 1964 (blue) chromaticity
β= (4) locuses. The full curve is the convex boundary of the
Ywr two locuses. In order to reduce the complexity for
Taking into account that D is the adaptation the optimization, the points on the convex boundary
factor between 0 and 1, and that D=1 is a full were used for the constraints (eq. 8).
adaptation and D=0 it has no adaptation:
⎛1 ⎞ 0.9
⎜ ⎟
Λ (0) = ⎜ 1 ⎟ , 0.8
⎜ 1 ⎟⎠
⎝ 0.7
0.6
⎛ R wr ⎞ 0.5
⎜β ⎟
y
⎜ Rw ⎟ 0.4
⎜ ⎟
G wr
Λ (1) = ⎜ β ⎟ 0.3
(5)
⎜ Gw ⎟
⎜ ⎟ 0.2
⎜ β
Bwr ⎟
⎜ Bw ⎟⎠
0.1
⎝
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
We can express the matrix gamma by this x
equation: Fig.2. Chromaticity coordinates (circles) for 82

illuminants. The dotted lines are the CIE1931 (black) and
Λ ( D) = DΛ (1) + (1 − D)Λ (0) (6) 1964 chromaticity locus. The full curve is the convex
boundary of the two standard observers.
Thus, from these equations (3-6) we can
express the CAT02 like this:
3. Results
sc ( D) = M −1Λ ( D) Ms = M −1[ DΛ (1) + (1 − D)Λ (0)]Ms
The matrix finally found following the previous
= DM −1 Λ (1) Ms + (1 − D) s methodology is:
(7)
⎛1.007245 0.011136 - 0.018381⎞
Since (1-D)s is equal or greater than 0 for any ⎜ ⎟
M new ⎜ - 0.318061 1.314589 0.003471 ⎟ (11)
object colour, the corresponding colour will be ⎜ ⎟
positive if this factor is equal or greater than 0. So ⎝ 0 0 1 ⎠
this leads to a constraint:
Note that the last row of the matrix is the same
s c (1) = M −1Λ (1) Ms ≥ 0 (8) as the last row of the matrix (eq. 2) suggested by
Brill and Süsstrunk5.
Therefore, the matrix M should be chosen so The next step is doing the performance
that for any point s on the CIE chromaticity locus evaluation with the new matrix. For testing the
inequality (8) holds. CAT02 we use four different matrices: the original
On the other hand, the matrix M should be CAT02 matrix (M02), the new matrix (Mnew), the BS
chosen to fit the experimental data well. Seven data matrix (MBS), and the HPE matrix (MHPE) given by
sets were used for deriving the new matrix: Colour equation below2:
Science Association of Japan (CSAJ), Kuo & Luo,
Lam and Rigg, Helson et al., LUTCHI, Breneman,
Braun & Fairchild, and McCann8. For measure the

⎛ 0.38971 0.68898 − 0.07868 ⎞ overall mean colour difference. The CAT02 with
⎜ ⎟ MBS becomes slight worse compared with the
M HPE = ⎜ − 0.22981 1.1834 0.04641 ⎟ (12)
⎜ 0 ⎟ CAT02 with M02. The CAT02 with the MHPE ranks
⎝ 0 1 ⎠ the third. The CAT02 with the Mnew performs the
Firstly, the visual experimental data sets were worst, which can also be expected since more
used for testing the CAT02 performance with constraints were added when the matrix was
different matrix. The average CIELAB colour optimised.
differences are listed in Table I. There are 21 data Finally, the CAT02 with each of the four
sets from seven groups of data. The last row lists the matrices was tested using the optimal colours under
mean colour difference values of all 21 data sets. the 81 test illuminants to check the non-negativity of
the corresponding colour. CIE illuminant D65 was
TABLE I used as the reference illuminant. Under the 81
Average colour difference (CIELAB) values under each illuminants, total number of optimal colours
data set for the CAT02 with M02, MBS, MHPE, and Mnew. (TNOC) generated is 20,252,868. The test results are
listed in the Table II, where NOI means the number
Matrix for the CAT02
of illuminants under which the CAT02 predicted
Data Sets M02 MBS MHPE Mnew corresponding colours with negative tristimulus
values, and where the NOCF means the number of
csaj.da.dat 5.36 5.57 6.33 6.59
optimal colours which failed the CAT02. The
kuo.da.dat 6.94 6.83 7.94 7.36 CAT02 with the original matrix failed to predict
corresponding colours for about 13.37 percent of the
kuo.dt.dat 3.57 3.68 4.57 3.89 optimal colours. The CAT02 with MBS has a
lam.da.dat 5.61 6.06 7.18 8.06 problem for about 3.45 percent of the optimal
colours. For the CAT02 with MHPE, the CAT02
helson.da.dat 7.12 7.45 7.99 9.39 predicted negative corresponding colours for about
lutchi.da.dat 7.92 7.93 7.92 10.34 0.2 percent optimal colours. While for the CAT02
with the new matrix, it successfully predicted
lutchi.dd.dat 6.88 6.88 6.53 7.76 corresponding colours for all the optimal colours.
lutchi.dw.dat 7.46 7.85 9.14 7.63 According to the non-negativity test for the
tristimulus values of the corresponding colours, the
brene.p1.dat 7.09 7.4 6.98 10.11 best is the CAT02 with the new matrix. The results
further confirmed our mathematical approach for
brene.p2.dat 5.3 5.95 4.97 8.23
searching the chromatic adaptation matrix for the
brene.p3.dat 7.85 8.53 9.29 9.63 CAT02 is correct.
brene.p4.dat 9.99 10.83 11.74 12.6
TABLE II
brene.p6.dat 8.95 9.31 8.36 10.78 Number of illuminants (NOI) under which the CAT02
brene.p8.dat 6.41 7.6 8.86 10.29 predicted negative corresponding colours and number of
optimal colours (NOCF).
brene.p9.dat 15.51 16.19 17.44 17.79 CAT02 with different matrix
brene.p11.dat 5.29 5 3.68 7.25 Original BS HPE New
brene.p12.dat 5.64 5.76 5.04 6.43 NOI 62 61 38 0
RIT1.dat 2.92 2.93 3.48 3.8 NOCF (%) 13.37 3.45 0.2 0
RIT2.dat 5.05 5.04 5.46 5.18
RIT3.dat 4.19 4.39 6.01 5 4. Conclusions
RIT4.dat 3.38 3.26 3.69 4.15 A mathematical approach has been proposed for
searching the chromatic adaptation matrix for the
Mean 6.59 6.88 7.23 8.20
CAT02. This approach combines the non-negativity
constraint for the corresponding colours’ tristimulus
values with the minimisation of the colour
As expected, the CAT02 with its original
differences between the tristimulus values of the
matrix performs the best since the matrix was
corresponding colours obtained by visual
derived based on the 21 data sets via minimising the
observations and tristimulus values of the

corresponding colours predicted by the model,

which resulted in a constrained non-linear
optimisation problem. For fitting the experimental
results, all the data sets used to derive the original
matrix have been used again here. A new matrix has
been found using the MATLAB routine “fmincon ”.
The performances of the CAT02 with the M02,
MBS, MHPE and Mnew have been tested. According to
the test for fitting the visual results, the ranking
order from the best to the worst is the CAT02 with
the M02, MBS, MHPE and Mnew respectively. While
according to the non-negativity test using the
optimal colours under 81 illuminants, the ranking
order is exactly the reverse order of the fitting visual
result test. The CAT02 with the Mnew is the only one
combination which successfully predicted
corresponding colours for all optimal. Thus it seems
that we have to scarify the accuracy in order to
satisfy the non-negativity constraint for the
corresponding colours.
It was thought that the CIECAM02 problem is
also solved by replacing the original CAT02 matrix
by the new one. Unfortunately, it was found that
some of the optimal colours also failed the
CIECAM02. It was also found that the failure of the
CIECAM02 using the optimum colours has a
relation with the correlated colour temperature
(CCT) of the test illuminant. Further work toward
repairing the CIECAM02 is under way.
Acknowledgements
This research was supported by the Ministerio de
Educación y Ciencia (Spain) under the grant
DPI2005-08999-C02-02. Esther Perales would like
to thank the Spanish Ministry for Education and
Science for the PhD grant (BES-2006-13518) that
she has received.

Spectral and photometric analysis of textured surfaces
C. Plata, J. L. Nieves, E. M. Valero, J. Romero
Departamento de Óptica. Facultad de Ciencias. Universidad de Granada.

Campus de Fuentenueva, Ed. Mecenas, 18071 Granada.
Email: cplata@ugr.es
ABSTRACT:
In the present work it is presented a method’s combination (four-source photometric stereo and
pseudo-inverse method) that allow recovering surface information (normal vectors and albedo) of
a sample and spectral information (reflectance) in each pixel. This information is used to simulate
a set of samples under different directions of illumination in a first step, and under different
geometries of illumination and different illumination sources in a second step.
Keywords: Photometric stereo, reflectance.

[1] R. J. Woodham, Reflectance map techniques for analyzing surface defects in metal castings. Technical
Report AI-TR-457, MIT A.I. Laboratory, (1987).
[2] R. J. Woodman, “Photometric method for determining surface orientation from multiple images” Optical
Engineering, 19, 139-144 (1980).
[3] K. Ikeuchi, “Determining surface orientations of specular surfaces by using the photometric stereo
method”, IEEE Transactions on Pattern Analysis and Machine Intelligence, 3, 141-184 (1981)
[4] G. McGunnigle, M. Chantler, “Rouge surface description using photometric stereo”, Measurement Science
and Technology, 699-709 (2003).
[5] M. S. Drew, “Photometric stereo without multiple images”, Human Vision and Electronic Imaging, 3016,
369-380 (1997).
[6] B. A. Wandell, “The synthesis and analysis of color images”, IEEE Transactions on Pattern Analysis and
Machine Intelligence, 9, 2-13 (1987).
[7] J. Hardeberg, F. Schmidtt, H. Brettel, “Multispectral color image capture using a liquid crystal tunable
filter”, Optical Engineering, 41, 2532-2548 (2002).
[8] F. H. Imai, R. Berns, “Spectral estimation using trichromatic digital cameras”, International Symposium on
Multispectral Imaging and Color Reproduction for Digital Archives, 42-49 (1999).
[9] C. Plata, J. L. Nieves, J. Romero, “Combining spectral and photometric stereo techniques for reflectance
estimation using an RGB digital camera”, in Proceedings of CGIV’08, 516-518 (2008).
[10] S. Barsky, M. Petrou, “The 4-source photometric stereo technique for 3-dimensional surfaces in the
presence of highlights and shadows”, IEEE Transactions on Pattern Analysis and Machine Intelligence, 25,
1239-1252 (2003).
[11] E. M. Valero, J. L. Nieves, S. M. C. Nascimento, K. Amano, D. H. Foster, “Recovering spectral data from
natural scenes with an RGB digital camera and colored filters”, Col. Res. and Appl., 32, 352-360, 2007.

[12] X. Zhang, B. A. Wandell, “A spatial extension of cielab for digital color image reproduction”, Soc. For
Info. Disp. Symp. Tech. Digest, 27, 731-734 (1996).
1. Introduction and objectives 2. Fundamentals

Photometric stereo techniques has been long used in 2.a. Photometric Stereo Fundamentals
order to recover 3D information about surfaces.
Let us consider a Lambertian surface patch
These methods can be classified in what Woodham1
with albedo ρ and normal N. Albedo is defined as
refers to as direct methods and indirect methods. The
the fraction of the incident light reflected by a
first methods are those which try to measure
surface. If the surface patch is lit by three light
distances ranges directly, while the indirect methods
sources with directions L1, L2 and L3, the intensities
attempt to determine distances by measuring
Ik of the obtained pixels can be expressed as1:
parameters calculated from images of the
illuminated objects. Shape from photometric stereo Ik = ρ(Lk·N) (1)
was conceived by Woodham in the early eighties
where k = 1, 2, 3 represents the illuminant directions
and has been extensively studied with both
and (·) the scalar product of two vectors. The pixel
theoretical and experimental approaches2-5. All
intensities can be stacked to obtain the pixel
approaches published since then can be classified
intensity vector I = (I1, I2, I3)T. The light vectors can
according to the assumptions the authors make about
also be stacked row-wise to form the illumination
the surface they are dealing with and the type of
matrix [L] = (L1, L2, L3)T. Then, equation (1) could
problem they want to solve.
be rewritten in matrix form:
Object’s colour depends on the registered
I = ρ[L]N (2)
colour signal6, that’s the reason why sometimes is
more useful to obtain the spectral reflectance for If the three light directions Lk do not lie on the
complete object characterization, because this same plane, matrix [L] is non-singular and can be
function is independent of illumination. inverted, giving:
Multispectral imaging can recover spectral radiance
[L]-1I = ρN (3)
or reflectance for each pixel of a scene of interest7.
Usually a multispectral system consist on an RGB or Since N has unit length, both the normal (as the
monochrome digital camera coupled with a number direction of the obtained vector) and albedo (as its
of wide-band or narrow-band colour filters8. As length) can be recovered.
opposing to conventional imaging devices, they
As can be seen we only need three images to
capture illuminant-independent images and allow
apply photometric stereo. In this work we have used
accurate spectral and colorimetric reproduction of
the so called four source photometric stereo10, where
color images for any lighting conditions.
four sources are used instead of three. Photometric
When we are dealing with real objects, we find stereo techniques are based on the constraint of
that most of them show textures. The appearance of Lambertian surfaces, what means that the algorithm
this kind of objects will depend on the direction of will fail in presence of highlights. As we are
illumination, and this will be a problem if we want working with real images we cannot be sure about
to recover reflectance from them9. The same pixel our samples are Lambertian, so taking captures
will give a different camera response under different under four sources allows us to select the best set of
illumination directions although the illuminant is the three in the albedo and normal recovery process, that
same. In the present work we propose a combination is, we can eliminate the one containing a highlight.
of photometric stereo and multispectral techniques
2.b.- Spectral reflectance estimation
to avoid the problem of reflectance recovery from
textured surfaces. This combination will allow us to In the so-called “pseudo-inverse method”11,
simulate a textured sample under different directions given a set of training spectra S (which can be
of illumination and under different illuminants. spectral radiance of reflectance) and the
corresponding set of experimental camera responses
q, a recovery transformation matrix D is defined by:
D = Sq+ (4)

where q+ is the pseudo-inverse of q. If q has full images under the same directions of illumination in
rank, then q+ = (qTq)qT, where qT is the transpose of order to compare the simulated images with the
q. An estimation set of spectra S1 may then be original, 24 images under 4 different directions of
obtained from the corresponding set of camera illumination makes a total number of 96 test images.
responses q1 by applying the transformation D, that
is:
S1 = Dq1 (5) DigiLite
1
Using this method, there is no need to use any
mathematical bases or to know the spectral 0.8
normalized SPD
sensitivity of the camera sensors.
0.6
2.c.- Computations
Two different simulations have been made in 0.4
this work. The first one uses the information
recovered with the photometric stereo method in 0.2
order to simulate the sample under different
directions of illumination. Once we have recovered 0
400 450 500 550 600 650 700
albedo and normal, we can consider equation (1) wavelength (nm)
with only one direction of illumination: Fig.1. Normalized SPD of the fluorescent lamp used in the
first set of simulations.
I = ρ(L·N) (6)
That way, using just one light direction we can
Two metrics were used to check the accuracy
simulate the intensity in each point of the sample
of results: root mean square error (RMS) over RGB
under any direction of illumination.
differences between the two images, and
The second application extends the previous SCIELAB12 differences.
simulation to a more general case. It is possible to
Results over the 96 images can be seen in
model albedo in each pixel as:
Table I. The first row shows several statistics over
ρ =∑E(λ)S(λ)Q(λ) (7) RMS differences and the second row the same
statistics over SCIELAB differences. Since RGB
where E(λ) is the spectrum of the illuminant, S(λ) is
values goes from 0 up to 65536 our RMS results are
the reflectance of that pixel and Q(λ) represents the
quite good. SCIELAB differences are very low too
camera’s sensitivity for each channel. Starting on the
with the mean color differences under 5 units.
results from the photometric stereo step, it is
possible to use albedo to recover reflectance9. Once
we have reflectance in each pixel of the sample, we TABLE I
only need the spectrum of one illuminant and the Results for application I.
camera’s sensitivities in each channel to simulate the Std Perc Perc
Max Min Mean
sample not only under any direction of illumination, Dev 95 99
but under any illuminant too. RMS 315 0 27 21 66 102
SCIELAB 65.47 0.03 4.27 3.57 11.54 18.46
3. Method Fig. 2 shows an example for one of the samples

Images were captured with a CCD camera Retiga (Gris 1390) captured and simulated under 270º. In
1300 from QImaging. As test set, we used 24 the figure, the first row presents three images. The
samples of frontage’s cover captured under four first one is the original image captured by the
different directions of illumination: 0º, 90º, 180º and camera, and the second image is the simulated one.
270º. The images had a size of 100x100 pixels. As it can be seen, in this case is very difficult to find
visual differences between the original and the
3.a. Application I
simulated images. The third image is the original
In this first set of simulations, all samples were image with pixels with SCIELAB differences higher
taken with the mentioned directions using as than 10 units painted in green. In the second line,
illuminant a fluorescent lamp (Digilite) which there are two histograms. The first histogram shows
normalized spectral power distribution (SPD) can be the RMS differences over all the pixels on the
seen in Fig. 1. From those four images, albedo and image, and the second one the SCIELAB differences
normals were recovered and used to simulate the over all pixels too.

Of course results in this case are worse than

before, but they can be considered acceptable from a
colorimetric point of view. Fig. 4 shows one
example of this application where Alabastro1384 is
captured and simulated under 0º and the
Original Simulated
incandescent lamp. Again, in the first row are shown
Gris 1390 Gris 1390
the original image, the simulated one and the
1000 6000
original image with those pixels with SCIELAB
5000
800
differences higher than 10 painted in green. In the
4000
600 second line there are two histograms: the first one
3000
400 for RMS differences and the second one for
2000
200
SCIELAB differences. In both histograms peaks are
1000
displaced to the right, what means that differences
0 0
0 5 10 15 20 25 30 35 40 45 50 55 60
RMS Diferences
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
SCIELab differences are higher than before, but even in this case original
Fig.2. Example of simulation I: Gris 1390 simulated and and simulated images are quite similar.
captured under 270º.
3.b. Application II
In the second set of simulations, we started
again from normals and albedo recovered from the
same set of images as before. But now, we recovered Original Simulated
reflectance from albedo for each one of the 24 1200
Alabastro 1384, 90º, Incandescent
3000
Alabastro 1384, 0º, Incand
images and simulated the albedo under three 1000 2500
different lamps, two different fluorescents and one 800 2000
incandescent, which normalized SPDs can be seen in 600 1500
Fig. 3. The samples were captured too under those 400 1000
lamps in order to compare. In this case, the 24 200 500
samples were captured and simulated again under 0

0 500 1000 1500 2000
0
0 10 20 30 40 50 60
the same four directions of illumination and under RMS differences SCIELab differences
the three illumination sources, so 288 images were Fig.4. Example of simulation II: Alabastro 1384 simulated
compared. and captured under 0º and incandescent lamp.
1
Digilite 4. Conclusions
0.8 Incandescent
Tri-Lite In conclusion, we have shown that four-source
normalized SPD
0.6
photometric stereo allows a simple 3D modelling of
surface properties under Lambertian assumption. We
0.4 have developed and efficient method to simulate
several real-world textured objects under different
0.2 illumination geometries. Finally it is possible to use
albedo in spectral analysis to recover spectral
0
400 450 500 550 600 650 700
reflectances from linear pseudo-inverse.
wavelength (nm)
Fig.3. Normalized SPD of the lamps used in the second set
of simulations. Acknowledgements
This work was supported by Spanish Ministry of
Same metrics as before were used to analyze results. Education and Science and FEDER through the
Global statistics can be seen in Table II. grant number FIS2007-60736.
TABLE II
Results for application II.
Std Perc Perc
Max Min Mean
Dev 95 99
RMS 9285 1 588 884 2564 4486
SCIELAB 244.44 0.05 18.96 15.60 52.19 74.15

¿Son temporalmente estables las

fuentes luminosas de las cabinas de iluminación?
R. Roa, R. Huertas, L. Gómez-Robledo, M. Melgosa
Departamento de Óptica. Facultad de Ciencias (Edificio Mecenas).

Campus de Fuentenueva, Universidad de Granada, 18071 Granada (España).
Email: rafaroa@ugr.es
RESUMEN:
El objetivo de este trabajo es estudiar la evolución temporal de distintas fuentes de iluminación
disponibles en cabinas de iluminación para evaluar cómo varían sus acuerdos con los iluminantes
a los que representan. Para ello se emplean parámetros como la temperatura correlacionada de
color, el índice CIE de rendimiento en color, distintas métricas espectrales y las diferencias de
color CIEDE2000 entre las muestras de la carta GretagMacbeth ColorChecker iluminadas por la
fuente simuladora o por su iluminante asociado. Se concluye que, cuando necesitamos realizar
medidas colorimétricas de alta precisión, el estado de la fuente debería ser conocido, ya que es
temporalmente dependiente y no siempre se puede asumir que sea idéntico a un iluminante CIE.
Palabras clave: Fuentes de luz, iluminantes, CCT, CRI, CSCM, SCI, RMSE, WRMSE,
CIEDE2000.

[1] CIE 15:2004, 3rd Edition, Colorimetry. Technical report, CIE Central Bureau, Vienna (2004).
[2] R. Huertas, M. J. Rivas, M. Melgosa, M. Sánchez-Marañón, S. Bhosle, J. J. Damelincourt, “Uniformirty of
lighting in color assessment cabinets”, in SPIE Proceeding of 19th Congress of the International
Commission for Optics (ICO XIX), Volume 4829, 879-880, Firenze, Italy (2002).
[3] R. Huertas, R. Roa, M. A. López-Álvarez, L. Robledo, M. Melgosa, “Comparación entre fuentes
simuladoras e iluminantes”, en Libro de Actas del VII Congreso Nacional de Color, 37-38, Madrid, España
(2007).
[4] R. Roa, R. Huertas, M. A. López-Álvarez, L. Gómez-Robledo, M. Melgosa, “Comparación entre
iluminantes y fuentes simuladoras”, Opt. Pura Apl., 41 (3) 291-300 (2008).
[5] R. Roa, R. Huertas, L. Gómez-Robledo, M. Melgosa, “Temporal evolution of light sources available in
comercial color cabinets”, in Proceedings of 4th Balkan Conference on Lighting – Balkan Light 2008, 271-
278, Ljubljana, Slovenia (2008).
[6] J. Hernández-Andrés, R. L. Lee Jr., J. Romero, “Calculating correlated color temperatures across the entire
gamut of daylight and skylight chromaticities”, Appl. Opt., 38, 5703-5709 (1999).
[7] CIE Publication 13.3-1995, Method of measuring and specifying colour rendering properties of light
sources. Technical report, CIE Central Bureau, Vienna (1995).
[8] CIE Publication 142-2001, Improvement to industrial colour-difference evaluation. Technical report, CIE
Central Bureau, Vienna (2001).

1. Introducción GretagMacbeth Spectralight III y VeriVide CAC 60.

En primera aproximación, el color percibido de un Las medidas de la radiancia de las fuentes luminosas
objeto depende de la fuente de iluminación o fueron realizadas sobre un blanco de referencia
iluminante escogido, de la reflectancia espectral del PTFE empleando el espectrofotómetro Photo
propio objeto y del Observador Estándar empleado1. Research SpectraScan PR-704. Dichas medidas se
Para el observador, la Comisión Internacional de realizaron a las siguientes horas de uso de las
Iluminación (CIE) recomienda el uso de los fuentes: 0, 0.5, 1, 2, 3, 4, 5, 10, 20, 30, 40, 50, 75 y
Observadores Estándar 1931 ó 1964, dependiendo 100 horas.
del ángulo subtendido por la muestra. Estos Se ha estudiado la evolución del acuerdo
observadores están definidos por sus respectivas iluminante – fuente simuladora empleando las
funciones de mezcla. Históricamente, como fuentes siguientes métricas para la comparación de valores
de iluminación, la CIE propone los siguientes espectrales: RMSE (Root Mean Square Error),
iluminantes: A, C, D50, D55, D65 y D75. WRMSE (Weighted Root Mean Square Error), SCI
Recientemente se han añadido otros iluminantes que (Spectral Comparison Index) y CSCM (Colorimetric
representan lámparas fluorescentes (iluminantes FL1 and Spectral Combined Metric). Una completa
a FL12). Actualmente, la CIE también proporciona descripción de estas métricas se puede encontrar en
las distribuciones espectrales de nuevas lámparas el anexo final de la referencia5. También se ha
fluorescentes (FL3.1 a FL3.15) y de lámparas de calculado la temperatura correlacionada de color
descarga de alta presión (HP1 a HP5)1. (CCT)6 de las fuentes de iluminación y su índice
Trabajos anteriores nos han mostrado que CIE de rendimiento en color (CRI)7. Finalmente se
existe una importante ausencia de uniformidad en las han obtenido las diferencias de color CIEDE20008
cabinas de iluminación comerciales2. Además, entre los chips de la carta GretagMacbeth Color
estudios previos nos indican que hay diferencias Checker iluminada por las fuentes simuladoras o por
relevantes entre los iluminantes y sus fuentes sus respectivos iluminantes.
simuladoras3,4. Otros trabajos indican que es
importante conocer el estado de la fuente de
iluminación que estamos empleando, ya que es 3. Resultados y discusión
temporalmente dependiente5. El objetivo de este En la tabla I se presentan los valores para las
trabajo es profundizar en este último punto y por distintas métricas. En la primera columna se muestra
tanto estudiar la evolución temporal de distintas el promedio entre los distintos valores de las
fuentes de cabinas de iluminación comerciales. métricas a las diferentes horas de uso. En la segunda
aparecen las desviaciones estándar de estos
promedios y los coeficientes de variación entre
2. Método paréntesis. En la tercera y cuarta columnas se
Se han estudiado tres tipos de fuentes de presentan los valores máximos y mínimos para cada
iluminación: D65, A o F (según la cabina) y TL84, métrica y las horas a las que las fuentes alcanzan
disponibles en dos cabinas de iluminación: dichos valores.
TABLA I
Métricas para el análisis de datos espectrales.
RMSE WRMSE
Desv. Est. Desv. Est.
Fuente Promedio Máx Mín Promedio Máx Mín
(CV) (CV)
SLIII 0.0029 0.1364 0.1210 0.0005 0.0135 0.0110
D65
0.1315 0.0121
(2.2%) (10h) (100h) (4.1%) (0h) (100h)
SLIII 0.0012 0.0440 0.0397 0.0001 0.0042 0.0037
A
0.0419 0.0040
(2.9%) (75h) (2h) (2.8%) (75h) (2h)
SLIII 0.0103 0.2042 0.1642 0.0005 0.0314 0.0296
TL84
0.1712 0.0306
(6.0%) (0h) (0.5h) (1.8%) (100h) (0h)
CAC 60 0.0087 0.4078 0.3738 0.0030 0.0549 0.0434
D65
0.3796 0.0450
(2.3%) (0h) (1h) (6.6%) (0h) (1h)
CAC 60 0.0037 0.0909 0.0768 0.0005 0.0122 0.0102
F
0.0841 0.0113
(4.4%) (100h) (0h) (4.8%) (100h) (0h)
CAC 60 0.0071 0.1905 0.1615 0.0016 0.0338 0.0275
TL84
0.1843 0.0325
(3.8%) (100h) (0h) (4.8%) (100h) (0h)

SCI CSCM
Desv. Est. Desv. Est.
Fuente Promedio Máx Mín Promedio Máx Mín
(CV) (CV)
SLIII 0.0016 0.0519 0.0439 7.36
D65
0.0473 8.81 0.51 (5.8%) 9.78 (0h)
(3.3%) (0h) (100h) (100h)
SLIII 0.0003 0.0101 0.0090
A
0.0095 9.34 0.40 (4.3%) 9.88 (2h) 8.51 (0h)
(3.1%) (75h) (2h)
SLIII 0.0039 0.1349 0.1220 26.16 18.84
TL84
0.1297 24.84 1.90 (7.7%)
(3.0%) (100h) (1h) (5h) (0h)
CAC 60 0.0114 0.2169 0.1728 22.42 19.40
D65
0.1795 21.16 0.82 (3.9%)
(6.4%) (0h) (1h) (100h) (0h)
CAC 60 0.0012 0.0283 0.0239 32.65 28.88
F
0.0262 30.82 1.01 (3.3%)
(4.5%) (100h) (0h) (100h) (0h)
CAC 60 0.0051 0.1434 0.1210 26.02 15.18
TL84
0.1373 24.17 2.79 (11.6%)
(3.7%) (100h) (0h) (100h) (0h)
En la tabla se observa que los resultados En la tabla III se muestran los cálculos de CRI.
obtenidos para las cuatro métricas no son iguales. Las fuentes A de la cabina Spectralight III y TL84
Esto se debe a que cada métrica está basada en de la CAC 60 son las que sufren las menores
distintas propiedades de las distribuciones variaciones en CRI con el envejecimiento. Las
espectrales de las fuentes. También se observa que variaciones de aproximadamente una unidad en este
los valores crecen ligeramente con el tiempo (esto índice pueden considerarse no significativas para
es, el acuerdo iluminante – fuente simuladora nuestros propósitos. Como las variaciones que
empeora con el tiempo), excepto para la fuente D65 hemos obtenido son de este orden o menores,
de la cabina Spectralight III, en la cual los valores de podemos concluir que no son significativas.
las métricas disminuyen con el envejecimiento de la
fuente. TABLA III
En la tabla II se muestran los resultados de los Evolución temporal del índice CIE de rendimiento en
cálculos para la CCT. Puede verse como la fuente A color.
de la cabina Spectralight III es la que tiene un menor CRI
coeficiente de variación. Por el contrario, la fuente Desv. Est.
Fuente Promedio Máx Mín
D65 de la cabina CAC 60 es la que tiene el (CV)
coeficiente de variación más grande. La fuente D65 SLIII 0.44 95
D65
95 94 (0h)
de la cabina Spectralight III es la única cuya CCT (0.5%) (20h)
aumenta con el envejecimiento de la fuente. La SLIII 0.05
A
98 98 (2h) 97 (0h)
fuente D65 sufre una gran disminución en la CCT (0.1%)
con el envejecimiento. SLIII 0.27
TL84
97 98 (0h) 96 (1h)
(0.3%)
CAC 60 0.44 94
TABLA II D65
95 96 (0h)
(0.5%) (100h)
Evolución de la temperatura correlacionada de color. CAC 60 0.44 91
F
91 92 (0h)
CCT (K) (0.5%) (100h)
Desv. CAC 60 0.09
Fuente Promedio
Est. (CV)
Máx Mín
TL84
97 97 (0h) 97 (75h)
(0.1%)
SLIII 6129 36 6263 6084
D65 (6500) (0.6%) (100h) (2h)
SLIII 2823 2833 2814 En la tabla IV se presentan las desviaciones
A
6 (0.2%) estándar de las diferencias de color entre iluminante
(2856) (0h) (2h)
SLIII 3698 30 3761 3663 y fuente simuladora a lo largo del tiempo en
TL84 (4000) (0.8%) (0.5h) (100h) unidades CIEDE2000. En la primera columna
CAC 60 6117 133 6539 5992 mostramos el promedio de estas desviaciones
D65 (6500) (2.2%) (0h) (100h) estándar para todos los chips a lo largo del tiempo
CAC 60 2549 13 2574 2527 para cada fuente. En la segunda se encuentra la
F (2856) (0.5%) (0h) (100h) máxima desviación estándar para cada fuente y el
CAC 60 3747 47 3883 3705 chip que la produce. En la tercera columna se
TL84 (4000) (1.2%) (0h) (100h) encuentra la mínima desviación estándar y su chip
asociado.

TABLA IV
Desviaciones estándar en las diferencias de color
CIEDE2000 para los chips de la carta GretagMacbeth
ColorChecker.
Chips y Desv. Est. en unidades CIEDE2000
Fuente Promedio Máx Mín
SLIII
D65
0.055 0.158 (13) 0.003 (24)
SLIII
A
0.012 0.030 (15) 0.000 (23)
SLIII
TL84
0.056 0.192 (13) 0.001 (24)
CAC 60
D65
0.091 0.409 (13) 0.002 (20)
CAC 60
F
0.039 0.118 (15) 0.002 (24)
CAC 60
TL84
0.120 0.437 (13) 0.000 (24)
Realizando un análisis completo, se encuentra

que los chips 8 y 13 tienen las mayores desviaciones
estándar en las diferencias de color entre iluminante
y fuente simuladora. Además, la fuente A de la
cabina Spectralight III es la que presenta las
menores variaciones en las diferencias de color con
el tiempo. Las mayores desviaciones estándar se
obtienen con la fuente TL84 de la cabina CAC 60.
Finalmente, puede verse como las muestras
acromáticas (chips 19 a 24) sufren las menores
variaciones con el envejecimiento de la fuente.
4. Conclusión
Este trabajo está relacionado con el trabajo actual
del Comité Técnico de la CIE TC1-44. Se observa
que la fuente A de la cabina Spectralight III es la que
sufre las menores variaciones con el envejecimiento,
atendiendo a la mayoría de los índices. La fuente
D65 de la cabina Spectralight III se aproxima mejor
a su iluminante asociado conforme envejece. Por el
contrario, la fuente D65 de la cabina CAC 60 sufre
un envejecimiento más acusado que el resto.
Podemos concluir que, cuando queremos
realizar medidas colorimétricas de alta precisión
(como experimentos sobre umbrales visuales), es
necesario conocer el estado de la fuente de
iluminación que estamos empleando, ya que es
temporalmente dependiente y no siempre podemos
asumir que es idéntica a un iluminante CIE.
Agradecimientos
Proyecto de Investigación FIS2007-64266,
Ministerio de Educación y Ciencia, España.


Ciencia y La Tecnologia Del Color Español

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Ciencia y La Tecnologia Del Color Español

Загружено:

Авторское право:

Доступные форматы

CIENCIA Y TECNOLOGÍA

Título: Ciencia y Tecnología del Color 2008. Workshop on “Colorimetry and

Editores: Joaquín Campos Acosta y Rafael Huertas Roa, Red Temática

Depósito Legal: GR-4028/2009

Edita e Imprime: Copicentro Granada S.L.

24-26 de Septiembre de 2008

En 2008 la reunión anual de la red “Ciencia y Tecnología del Color” ha adoptado la

24-26 de Septiembre de 2008

1. Evaluating a colour difference formula in a new multistage colour vision model. E.

2. Analysis of the modulation transfer function (MTF) spectral behaviour in detector

3. Color in virgin-olive-oil tasting glasses. L. Gómez-Robledo, M. Melgosa, F. J. Heredia,

4. Iris color evaluation through multispectral systems. J. Herrera, M. Vilaseca, J. Pujol,

5. Reproducción espectral de valores triestímulo mediante descripciones B-Spline:

6. Colorimetric properties of thermoplastic polymers coloured with nanopigments and

8. Spectral and photometric analysis of textured surfaces. C. Plata, J. L. Nieves, E. M.

9. ¿Son temporalmente estables las fuentes luminosas de las cabinas de iluminación? R.

24-26 de Septiembre de 2008

Evaluating a colour difference formula

E. Chorro (1), E. Perales (1), F. M. Martínez-Verdú (1),

1. Departamento de Óptica, Farmacología y Anatomía. Universidad de Alicante.

REFERENCES AND LINKS

24-26 de Septiembre de 2008

1. Introduction after a last transformation, yield the achromatic, red-

The multistage colour vision model, ATTD051-

24-26 de Septiembre de 2008

24-26 de Septiembre de 2008

3.b. Study of the Uniformity and comparison with

24-26 de Septiembre de 2008

Analysis of the modulation transfer function (MTF) spectral

A. Fernández-Oliveras, A. M. Pozo, M. Rubiño

Departamento de Óptica. Facultad de Ciencias. Universidad de Granada.

REFERENCES AND LINKS

24-26 de Septiembre de 2008

1. Introduction Therefore, to investigate the suitability of the

24-26 de Septiembre de 2008

24-26 de Septiembre de 2008

direction, corresponding to a Nyquist frequency of

24-26 de Septiembre de 2008

24-26 de Septiembre de 2008

by fitting experimental values to a second-order TABLA III

Within the range of the visible spectrum

24-26 de Septiembre de 2008

24-26 de Septiembre de 2008

Color in virgin-olive-oil tasting glasses

L. Gómez-Robledo (1), M. Melgosa (1), F. J. Heredia (2),

1. Departamento de Óptica. Facultad de Ciencias. Universidad de Granada. 18071 Granada, Spain.

REFERENCES AND LINKS

24-26 de Septiembre de 2008

1. Introduction We poured 200 ml of oil in each standard cup,

2. Materials and Methods

24-26 de Septiembre de 2008

24-26 de Septiembre de 2008

Iris color evaluation through multispectral systems

J. Herrera, M. Vilaseca, J. Pujol, M. Arjona, M. de Lasarte

REFERENCES AND LINKS

24-26 de Septiembre de 2008

1. Introduction cosmetic colored lenses in three different groups:

24-26 de Septiembre de 2008

prostheses samples. In the same way, color gamuts 2.b. Methods

2. Experimental setup and methods

24-26 de Septiembre de 2008

TABLE I colored contact lenses, helping us to conclude how

The automatic classification of the samples is

3. Results and comparison