MEETINGS FOR 2008-2009
Abstracts for 5 November 2008
| |
MEETINGS FOR 2008-2009 Abstracts for 5 November 2008 |
WD Wright &
MacDonald/Green Awards:
An Educational Exercise
Accuracy
of cross-media colour memory
Philip Henry, Vien
Cheung and Stephen
Westland
School
of Design, University of Leeds
This work is
concerned with the investigation of colour memory of users (designers,
printers, students, etc) in
cross-media situations, particularly, between physical (printed) and
digital
(on-screen) colours. The preliminary experiment reported here supports
findings
by other researchers that there are systematic shifts between the
memory
colours of observers and the target colours that are being memorised.
The
results also confirm our hypothesis that when users are asked to select
a
colour on screen to represent a memorised target (or, we argue, to
represent a
colour that they have imagined) there are additional systematic
colorimetric
shifts. Users tend to select colours that are much more saturated than
the
target. There are implications of this work for soft-proofing and the
use of
imaging software by designers.
Assessing
Tooth Gloss using
Digital Imaging
Wen Luo
School
of Design, University
of Leeds
The aim of this study
was to assess
tooth gloss by a digital-imaging method. A gonio-imaging system was
developed
to measure the gloss of human teeth in
vitro.
18 polarised and non-polarised
images were acquired around the specular angle at 9 positions for one
measurement. The gloss component was extracted and normalised to a
theoretical
standard; a 2-dimensional BRDF (Bi-directional Reflectance Distribution
Function) curve was built to describe the gloss profile of the subject.
A set of porcelain teeth was used to
test the repeatability of the system and several human lateral teeth
were
etched to check the system’s capability to detect gloss
changes. The shape of
the BRDF curve indicated the gloss level of the surface; additionally
the area
under the curve was calculated to give a quantitative value of gloss.
From the
results of the stability test and human-teeth etching experiments, the
gonio-imaging system can be considered to be suitable for building the
tooth
gloss profile and to provide a sensitive measure of changes in gloss.
The system was developed to evaluate
gloss of human teeth in vitro, but
it
could be modified for the measurement of patient teeth in
vivo. These findings have implications for the use of a
digital
camera in the assessment of gloss related to dental practices, such as
tooth
surface enhancement and tooth bleaching.
Elza Tantcheva, Vien
Cheung† and Stephen Westland†
Department
of History of Art, School of
Humanities, University of Sussex
†School
of Design, University
of Leeds
The aim of this
research was to record and compare the
colours used in works of art in a form which would provide an
unambiguous
record but which would also be more familiar to the scholar working in
an
art-historical field. In the latter context, importance is placed on
visually
perceived colour rather than on colorimetric notation. Two metrics for
similarity were used; one based on the closest spectral match and the
other
based on the closest colorimetric match. We suggest that the
colorimetric
matches are closer to the originals. The investigation is grounded in
the case
study of four post-Byzantine churches in Arbanassi, Bulgaria. Such
presentation
allows the comparison of colours from different sites by overcoming
problems linked
to colour vision, colour memory and colour reproduction in print. This
is the
first time that this method of colour comparison has been used in art
historical research connected to a Bulgarian historical site in
particular.
Iván
Marín-Franch
School
of Electrical and Electronic Engineering, The
University of Manchester
Trichromatic
images
of natural scenes contain information about the reflecting surfaces in
those
scenes. The present work was aimed at comparing, in an objective
manner, the
ability of different colour-vision models to extract that information,
here
quantified by measures taken from Shannon’s information
theory. Two information
measures were used, one of the information potentially available to the
eye and
other of the information actually retrieved in a colour-matching task.
Estimates of both information quantities were obtained computationally
from a
database of 50 hyperspectral images of rural and urban natural scenes
with
different daylight illuminant changes. It was found that the
information
available (identical for all colour spaces related by invertible
transformations) is very large, and that most of that information can
be
retrieved in colour matching by the use of linear models of cone-signal
interactions and post-receptoral von Kries adaptation. It was also
found that
the spaces defined by popular colour-vision models represent image data
with
differing degrees of redundancy. For most of these spaces, the two axes
associated with chromatic properties carried more information than the
axis
associated with achromatic properties.
Milena Vurro, Yazhu
Ling and Anya Hurlbert
Institute
of Neuroscience,
Newcastle University
Although colour
and luminance textures are both
visual properties of natural surfaces and key cues in the distinction
and
recognition of objects, they have mostly been studied separately. The
non-uniformity in both chromaticity and luminance of natural surfaces
is not
captured by traditional studies of colour perception, which typically
employ
stimuli of uniform colour and brightness. To quantify and characterise
surface
chromatic texture, we propose a transformation model from RGB device
dependent
values of a camera system to device independent values. We employ the
camera
characterization model to obtain the chromaticity distributions of
seven
familiar objects under three
illuminations: banana (in three different stages of
ripeness), carrot,
apple, clementine, plum, strawberry and cucumber. The distribution of
within-surface cone contrasts for a given object forms a distinct
signature in
three-dimensional cone-contrast space, which transforms predictably
under changes
in illumination. For many natural surfaces, the distribution is an
elongated
cluster whose vector direction in cone-contrast space remains roughly
constant
under illumination changes, provided the contrasts are calculated with
respect
to the illumination white point. This feature provides a surface
descriptor
that remains stable under adaptation to the illumination, thereby
potentially
mediating colour constancy. In a second stage analysis, we calculate
the convex
hull containing the chromaticity distribution of the object and
compared the
‘gamut’ of the object under different
illuminations. We compute the correlation
between gamuts of the objects under different illuminants. For a given
object,
the correlation between gamuts varies across illuminants, but it is
highest for
the same illuminant, and lower for different illuminants. The finding
suggests
that the visual system may use the observed gamut of a familiar object
under an
unknown light to recover the best-matching illuminant by comparing the
observed
gamut with stored ”memory gamuts”. This information
may be used to calibrate
colours of other unfamiliar objects in the scene. To test these
physical
proprieties of the objects, we design a psychophysical experiment that
allows
the subject to adjust the entire distribution of colours of a familiar
object
under three different light conditions. The stimuli were presented in
different
shape conditions: solid 3D, 2D filled outline of the object, 2D neutral
disc.
The result of the experiment will be presented in the talk.
Monika Hedrich,
Alexa I. Ruppertsberg and Marina
Bloj
Bradford
Optometry Colour and
Lighting Lab, School of Life Sciences, University of Bradford
Previous studies comparing colour constancy across
diverse illumination changes have drawn an inconclusive picture as it
is not
yet firmly established if typical illumination changes, which are
likely to
occur during a daily routine (e.g. change between daylight and
tungsten), lead
to higher levels of colour constancy than atypical ones.
Using a real surface matching
task we investigated if either (a) the nature of illumination change
(typical
vs atypical) or (b) the learning illuminant had an influence on
observers’
colour constancy performance. For (a) observers learned a real surface
colour
under daylight and matched it either under tungsten (typical change) or
a
purple illuminant (atypical change). For (b) learning took place either
under
the tungsten or the purple illuminant and matching was always performed
under
daylight.
For all four experiments
surface colours were either learned as part of a three-dimensional (3D)
or a
two-dimensional (2D) set-up. We chose a mixed design with the four
illumination
changes as between-subject factors and the two learning set-ups (3D/2D)
as
within-subject factors. In total, twenty-eight colour-normal observers
participated. After learning a real surface colour for 20 s, an
illumination
change occurred and observers adapted for 2 min to the new illuminant
before
matching, which was always done in a 2D set-up. Six different target
colours
were tested.
The results showed no
evidence that observers performed better for typical than atypical
illumination
changes nor that the learning illuminant had a significant effect on
observers’
performance. However, learning real surface colours as part of a 3D
set-up
significantly improved observers’ colour constancy
performance. We conclude
that human colour constancy is able to cope with a wide variety of
illuminant
changes and benefits from the additional cues available in a 3D set-up.
Luis Garcia-Suarez,
Alexa I. Ruppertsberg and
Marina Bloj
Optometry
Colour and
Lighting Lab, School of Life Sciences, University of
Bradford
Gradients (smooth spatial variations in luminance
and/or chromaticity) are all around us as shading or other illumination
phenomena. They provide cues to light-source positions, object shape
and the
spatial layout of scenes. However, not much is known about how the
visual
system processes gradients. We measured detection and discrimination
thresholds
for horizontal achromatic gradients.
The gradient stimuli had a
fixed size (4°). Its horizontal luminance profile was either
linear or
sinusoidal and was generated such that its mean luminance was constant
and
equiluminous with the surround (56 cd/m2). Only its contrast varied
from trial
to trial. We used a temporal 2-AFC with a QUEST procedure to determine
contrast
thresholds for four observers in detection and discrimination
experiments.
In the detection experiment,
the background was either uniform (UB) or nonuniform (NUB:
mosaic-squares of
different luminance). Detection thresholds for gradients against UB and
those
for a step stimulus of a control study (against UB or NUB) were all
identical.
However, detection thresholds for gradients against NUB were on average
4 times
higher, indicating that NUB effectively disrupts the edge cue at the
boundaries
of the gradients. Thresholds for sinusoidal gradients were
significantly lower
than for linear gradients.
In the discrimination
experiment, two conditions were tested using only NUB; observers
indicated
which interval contained a gradient that was stronger (or weaker) than
a
reference gradient of fixed contrast. Observers performed better in the
stronger condition by an average factor of 2. For this condition,
thresholds
for sinusoidal compared to linear gradients were lower by a factor of
2.
The differences in thresholds
found between sinusoidal and linear gradients suggest that the visual
system
uses the information within the gradient for detecting or
discriminating it.
The asymmetry found between discrimination conditions implies that
increments
and decrements might be processed differently.
The Colour Group of Great Britain - Homepage
Webmaster:
Last Updated 13 October 2008