Subject Groups
The 1931 Standard Eye
Wednesday 5 April 2006 at Institute of Physics, 76 Portland Place, London, W1N 4AA
Organised by the Printing, Packaging and Papermaking Group of the Institute of Physics and the Colour Group (Great Britain)
Programme
This meeting marked the 75th anniversary of the international ratification of the 1931 Standard Observer.
14:00: Welcome
14:05: From the Greeks to 1931: A brief history of Trichromacy
Andrew Hanson, National Physical Laboratory
Units of measurement and specification have always been around. In the forms we use today, it could be argued that 75 years ago, colour was one of the last units to be finally internationally ratified. The basic colour model, that any colour may be matched by three primaries (trichomacy), is not new and has roots in classical times. This talk will examine the story of our understanding of trichomacy, culminating in the experiments that underpin the majority of scientific metrics for colour today.
14:30: Colorimetry in colour reproduction
Phil Green, Colour Imaging Group, London College of Communication
Colorimetry is today at the heart of a colour reproduction system as it provides the basis for translating colour between all the different capture and reproduction devices that are found in the graphic arts. Topics will include the use of XYZ tristimulus values as a 'linear light' encoding, the properties of the chromaticity diagram in additive colour reproduction systems, and examples of the use of more perceptually uniform spaces in the visualisation of colour difference and colour gamuts. The importance of standardisation in measurement will be discussed, together with the limitations of colorimetry in predicting effects such as metamerism and fluorescence.
15:10: Granville Tea (and exhibition)
15:40: 1931 and the non-standard Observers
Boris Oicherman, Colour and Imaging Research Group, Department of Colour and Polymer Chemistry, University of Leeds.
The CIE1931 represents the colour matching properties of an average observer. In reality, however, such an observer does not exist, and every real observer has colour matching functions which differ to varying extent from the standard and from each other. The consequence of these variations is the phenomenon of observer metamerism: a pair of stimuli can match in colour to one observer, and mismatch to another. In our talk we will give an overview of the problem, of its practical implications, as well as an updated report on our research into the “Uncertainty of Colour Matching”. We will conclude with the discussion of the open questions and a view of possible future developments.
16:20: CIE Colorimetry: After 1931
Mike Pointer, National Physical Laboratory
This talk will describe the developments in CIE colorimetry following the recommendation in 1931 of the 2° colour matching functions and the CIE sources and illuminants A, B and C. These developments will include the various colour difference equations, the 10° observer, uniform chromaticity diagrams, the D illuminants and the recently recommended colour appearance model CIECAM02. In addition, some thoughts as to the future might be offered!
16:50: Adjournment
Meeting report by Andrew Hanson
The gathering was arranged to commemorate 75 years of the CIE1931 Standard Observer, was a joint meeting between the Institute of Physics' 3P's Group and the Colour Group, and was well attended.
The meeting began with Andrew Hanson (NPL) who outlined some of the history of colour matching and measurement. He began with the Greeks who believed that ‘light’ shone from the eye and ‘read’ objects similarly to touch, and that the ‘primary’ colours were light and dark. Andrew pondered that since colour derives from absence and presence of light (albeit throughout a wavelength range), and that the luminance pathway seems to contain a greater amount of visual information than colour, to an extent the Greeks were right. The work of several Arabian scholars such as Ibn al-Haitham turned about Greek philosophy, but these advances remained largely untranslated in the west and sympathy for the Greek primaries remained beyond Newton’s time. Newton’s main contributions to colour were arguably: that since any coloured light exiting a prism could not be split further, there was nothing ‘primarily‘ special about red, blue or yellow; and that a centre of gravity approach could be used to make colours from differently hued lights arranged in a hue circle. Maxwell’s experiments used a tri-colorimeter approach to match arbitrary colours with mixtures of three ‘primaries’. Maxwell showed that no set of ‘real’ primaries could match all colours (as any light wavelength excites more than one eye detector). In the 1920’s work standardising the human colour vision system was carried out in Germany and England, and the data adopted by the Optical Society of America. This caused a stir when the CIE looked set to adopt a different set of data (17 observers) collected by Stiles at NPL and Wright at Imperial College London. Andrew described the political and scientific machinations between the North Americans and British one week before the data was ratified internationally in September 1931.
In a ‘virtual’ contribution, Arthur Tarrant described the personalities and working conditions of Stiles and Wright. The latter was a researcher at ImperialCollege and had considerably less funding than Wright at the NPL. As such, Wright ground his own optics.
Phil Green (London College of Communication) explained how the colour reproduction uses tristimulus colorimetry, adaptation, source colour, encoding, colour management, gamut and image states. Phil described issues in both acquisition and outputting of coloured images, and described transforms for coding input and output devices to a common ‘non device specific colour space’. Excellent visual demos showed the process of encoding (including adaptation of dynamic range), and to illustrate colour appearance he showed an image of a girl on a horse, whose top appeared orange – although actually a general blue cast of the image was transforming the neutral smaller ‘top’ area to a more orange colour. Phil described how CIE1934 fitted into various colour transforms, and how the inherent non-uniformity of this colour space lead to the adoption of CIE LAB colour space, which has since been improved upon (though arguably not significantly) by CIE DE 2000. The Mac computer operating system OSX had a profile connection space colour model included, as does the forthcoming PC Vista operating system. This will significantly increase the ubiquity of the use for CIE1931. Looking ahead, the remaining challenges seem to be: representation of small and large colour differences, mapping between gamuts, dealing with fluorescence, metamerism in multispectral reproduction, image quality for preferred reproductions, and finally how to deal with images containing luminance values greater than the perfect reflecting diffuser.
To put some of his present research into context, Boris Oichermann (LeedsUniversity) quoted Robert Weale: “There is no such thing as an average eye”. In response to questions regarding observer data spread in a previous talk, variation in the evaluations of the Photometric observer had already been shown, and Boris presented individual data sets for the 17 observers of Stiles and Wright, and similar results for the 49 observers in the 1964 observer experiments. All these data showed considerable spread - not purely accounted for by experimental error. Boris described the components of the eye accounting for variations in vision: a +/- 25% variation in lens density (and yellowing with age), macular pigmentation varying by +/- 50% (affected by diet: nourished by spinach, egg yolk and orange pepper but deteriorated by smoking), photoreceptor peak wavelengths can differ by 6 nm. All these factors combine to produce observer metamerism, or put another way – uncertainty in colour matching. Nimeroff (1961) and Allen (1970) worked on a ‘standard deviation observer’, as did Alfin and Fairchild (1997). These results were expressed as variations in response for individuals and groups of people. At Leeds, Boris is obtaining 5 observer’s data using the Tarrant Visual Colorimeter, a telespectroradiometer, and different viewed media. The results appear to reproduce the variability seen in the 1964 Stiles and Birch colour matching function data, and are similar to those of Alfin and Fairchild. In a second experiment, observers adjust monitors to match the colour of painted samples. The monitors are CRT and LCD displays with similar gamuts but different spectral characteristics. The results of 11 observers, with 5 repetitions and 10 test colours have a typical spread of CIELAB 5.5 delta E units. In conclusion, there is a significant spread in the perception of colour. The CIE standard observer has been validated as giving a good average, but perhaps an ‘observer metamerism aware system’ would be of value.
Mike Pointer (NPL) described the history of colour measurement since the momentous meeting in September 1931 which ratified not only the Standard Observer, but also the colour co-ordinate system, standard sources A, B and C, and standard measurement geometries. The early workers used visual photometry, and it was in 1935 that photocells were first used in earnest in the Hardy recording spectrophotometer. 1944 saw the publication of Wright’s book ‘Measurement of Colour’ – considered a seminal work. The 1940’s also saw work tackling the non-uniformity issue of CIE1931, producing the CIE UV system which was improved in uniformity by a factor of three. The 1960’s were particularly busy – with the introduction of the daylight ‘D’ illuminants, the 10 degree observer, and an improved U*V* colour space. The latter was finally revised in 1976 into the CIELUV system, currently favoured by the lighting industry (as it has a reasonably uniform chromaticity diagram). At the same time, the CIELAB system was introduced – a ‘uniform’ surface colour system with no associated chromaticity diagram, involving cube root maths and a delta E colour difference metric which was reasonably good for all colours. Since colour difference is considered important, advances continued – with improved CIE colour difference equations published in 1994, then 2000 - the latter with a special hue rotation term included just to ‘get the colour of blue jeans right’. As long ago as 1957, W. D. Wright complained: “[CIE] XYZ doesn’t tell us what it looks like”. While colour specification ignores influences such as surround colour, level of illumination and illumination colour, colour appearance tackles these issues. In 1982, Hunt’s colour appearance model considered stimuli in context of (concentrically arranged) proximal, background and surround fields along with illumination level and colour. This evolved into 1997 CIECAM97, then CIECAM02.
Looking to the future, colour is not everything - we see texture, highlights, translucency and context. Deriving a comprehensive set of visual correlates describing total appearance might be considered the next big challenge and work at NPL is tackling this. There may still be some room for improvement of uniform colour spaces and colour differences based in appearance space. How perception varies with field size and age is also being investigated by Vienot and others. A scan of present CIE TC activities informs us of the active fields: 1-55 (new uniform colour space related to appearance); 1-56 (related to the work of the previous speaker), 1-36 tables data of the 2 degree cone fundamental response functions in terms of energy and quanta. We might imagine a future with a grand unified theory of vision which considers colour, gloss, translucency and texture, all as functions of illumination angle/colour/level and angle of view.
^ To the top ^