1. In many ways the CIE system of colour specification has been remarkably successful.
Almost all important applications of colour measurement use it, and the
basic system has survived unchanged for over 60 years.
The additions made since 1931
Have led to improvements in some respects,
but have not changed the basic principles of the system in any way.
Tanveer Ahmed 1

2. These stem basically from the limited objectives of the system,
rather than from a failure to meet those objectives.
The CIE tristimulus values for a sample are related to the colour of the
sample, but ignore other important features such as
surface texture, gloss and sheen.
Thus a gloss paint sample and a matt paint sample might have the same tristimulus
values, but obviously will not look the same.
Whether the colours of the two samples will look the same depends
critically on the geometrical arrangements for illuminating and viewing
them.

3. Only if the instrumental geometry of illumination and viewing conditions is
similar to
 that used for visual matching will the colours be seen to be close.
An instrument will always average out the light reflected from the area being
measured
 (typically a 2 cm diameter circle).
In judging a colour visually some sort of averaging takes place, but the observer is
always conscious of any non-uniformity over the area viewed.
Thus a matt paint surface, a woven textile surface and a pile fabric will always
look
different from one another,
but their measured tristimulus values could be the same.

4. Ignoring all features other than colour, the tristimulus values for a sample give only
a limited amount of information.
The tristimulus values tell us the amounts of three
imaginary primaries
which if additively mixed will give
1. the same colour as a surface
2. illuminated by a standard source
3. and viewed by a standard observer
4. using one of the standard geometries
5. (assuming that the instrument does correspond to the standard conditions).
It follows that the mixture of the CIE primaries would be unlikely to match the
surface if
it was illuminated by a different source
or if the ‘match’ was viewed by an individual observer
or if different illuminating or viewing geometry was used.

5. Obviously a degree of control can be exercised
over the source and the geometry.
If these are important, we must try to ensure that the instrumental conditions
correspond as closely as possible
 to those
 to be used
when viewing the object visually.
The only choice as far as the standard observer is concerned is
 whether to use the 1931 (2°)
observer or the 1964 (10°) observer.
Neither is likely to correspond closely to any individual observer, but either may
well correspond reasonably closely with the average judgement of real
observers,
bearing in mind that in many applications
the product is mass-produced
and will be seen by many different observers.

6. Usefullness of the CIE System standard observer usefulness
A full specification of a colour requires X, Y and Z values
 (or equivalent sets such as
x, y and Y or
 L*, a* and b*)
for several different illuminants.
The results are still valid
only for the standard observer and could be unsatisfactory for a real observer.
This should not be a problem in practice since the need is usually for
the colours to be acceptable
to a large number of potential customers;
the standard observer is probably a better guide to the population in general
than any one observer.
It often happens that one particular individual
(the head dyer or senior buyer for a chain store, for instance)
inspects fabrics, and problems may arise if that individual’s colour vision is
appreciably different from that of the standard observer. Problems will be most
severe for highly metameric pairs of samples.

7. Illuminant limitations
Strictly speaking, the tristimulus values tell us nothing about the colour of a
sample
although, as discussed above,
with experience we can make a reasonable estimate of
the colour from either
 X, Y and Z
or x, y and Y values.
It is then essential that the illuminant used for the measurements is known.
Chromaticity coordinates of
x = 0.314
and y = 0.331 correspond to a neutral colour
if derived from measurements under illuminant D65,
but to a blue colour if derived from measurements under illuminant A.

8. Matches to target limitations
In many applications the aim is to match a particular target, which might be defined
by a set of tristimulus values.
If we produce a sample and wish to test whether this matches the target, the
sample and target measurements must correspond to
 exactly the same conditions
 (illuminant, standard observer, illuminating and viewing
geometries, and, in practice, the same instrument).
If, for example, the sample really is a good match to the target
but the tristimulus values are measured using different
standard observers for the sample and target,
the resulting tristimulus values
 and chromaticityncoordinates would be appreciably different.
Again the sample and target might have different surface structures, such as
those of matt paints, gloss paints or pile carpets:
the tristimulus values could then be identical
but the surfaces would look appreciably different.
Whenever possible sample and target should have the same surface structure.

9. Same Material & same instrument / observer / illuminant limitations
It is usually important that the colours within one batch of fabric, and between
repeat batches,
should match closely.
In these cases the samples will of course be of the
same material
 and the same dyes or pigments will have been used.
It would be natural to measure all the samples using the same instrument
and (if a spectrophotometer was used) to calculate
the tristimulus values for the same standard observer and standard illuminant.
Under these conditions, if the tristimulus values for a sample are very close
to those for the standard, then
the sample will be a close visual match to the standard
for any normal observer viewing
under a light source roughly equivalent to the standard illuminant used to
calculate the tristimulus values.
(Exceptions are known: for example, the appearance of metallic paints depends
very much on the illuminating and viewing geometries.)
If the variation of appearance with, say, viewing angle is different for the sample and
the standard, they may not match visually even though the instrumental results
(obtained with a different viewing angle) indicate that they should.