Colour Theory: Understanding and Modelling Colour

Last updated: 16 December 2008
Published in: Digitising analogue media | Creating new digital media |
Tags: colour management | workflow |

Comments (1)

Colour is an area of digital imaging that can lead to considerable confusion. This document provides a detailed introduction to colour theory.

Contents

• Introduction
• Understanding Colour
  • Light
  • Substance
  • Human Perception
  • Additive and Subtractive Colour
  • Links to further reading
• Modelling Colour
  • Historical Models
  • Contemporary Models
  • Other approaches
  • Links to further reading

Introduction

This is one of two JISC Digital Media's Advice documents covering colour management. This document deals with colour theory, outlining some of the science and history that informs our colour management practice. It considers the "why" questions rather than the "how to" questions, which are addressed by JISC Digital Media's second advice document Colour Management in Practice.

First let's shatter a myth: colour is not a simple objective thing we can point to. It is the way we perceive the interaction of light on substances. None of the individual colours we name (red, purple, orange) exist in reality - they are categories we create in order to describe and control the experience we call 'colour'. As we will see, colour is the result of a number of processes and interactions - some of them natural, some technological and others historical or cultural.

Colour is a fascinating but complicated area of enquiry. All we can hope to do here is provide a simple introduction to the subject. In addition to these core colour papers, JISC Digital Media has included some links to other information available online.

Understanding Colour

As we've said, colour is not a simple objective thing that exists in the world. It is the interaction of light on a substance, perceived by a human being. So in order to understand and effectively manage colour it is useful to have a basic understanding of all three parts of this process: (1) the nature and quality of light; (2) the way light interacts with material substances; and (3) the way that interaction is perceived by the human eye and brain.

Light

Light is a form of energy known as electromagnetic radiation (EM). Electromagnetic radiation originates from the Sun, but can also be produced artificially. Although EM exhibits both wave and particle characteristics, it is most commonly described as waves and characterised by its wavelength (measured in nanometres, nm). At one end of the EM spectrum are the very long radio waves, some of them metres in length. Towards the other end are tiny x-rays and gamma rays, which have wavelengths smaller than a billionth of a metre.

As the diagram below suggests, the part of the electromagnetic spectrum that our eyes can actually detect ("visible light") is tiny, less than 1%. It stretches from about 780 nanometres in wavelength (i.e. 0.00000078 metres) down to 380nm.

Diagram 1. Electromagnetic radiation from the Sun

Pure white light is an equal mix of waves from all parts of this visible spectrum. However, in reality light is seldom such an even mix. Sometimes there are a greater number of the longer waves in the light, causing it to appear reddish. At other times there are more shorter waves, producing bluish light. The particular tinge a "white" light has is described as its colour temperature and is measured in degrees Kelvin (K). Here are some colour temperatures for common light sources:

Table 1. Light sources and their approximate Kelvin values
Candle	2000K
Sunrise or Sunset	2500K
Standard household light bulbs	3000K
Noon on a sunny day "Daylight"	5500K
Electronic flash	6000K
Overcast sky	7500K
Blue sky	12000K

It might seem counter-intuitive that bluish light has a higher temperature than reddish light (since culturally we associate red with heat and blue with cold) but bluer light does have more energy present and so is hotter. The Kelvin scale itself was determined by progressively heating a black object so that it glowed red, then white, then blue.

The temperature of a light source will affect the appearance of a scene or of coloured objects. Our eyes and brain tend to compensate for this, but when, for example, a more objective device like a camera captures a candle-lit scene, the resulting photographs will often appear much too orange (unless it has been calibrated to work with the candle-light).

Light Sources

Since light is a form of electromagnetic energy, anything that emits, re-emits or conducts energy in sufficient quantities will produce light. Light sources (or 'illuminants') generally produce light in one of three ways:

• Incandescence - where a solid or liquid is hot enough to emit light (e.g. the sun, a candle or the standard Tungsten light bulb found in most homes)
• Gas discharge - an electric current is passed through gas, producing light (e.g. a mercury lamp or neon light)
• Photoluminescence - phosphors absorb and then re-emit light, either concurrently (in which case the source is said to be 'fluorescent') or later on (this is known as 'phosphorescence').

A standard fluorescent tube uses a combination of gas discharge and photoluminescence. It is a phosphor-lined tube containing a mercury lamp. The lamp lights the phosphors, which then re-emit the light we see.

We've said that light sources can be characterised by their colour temperature. This is known as their Correlated Colour Temperature (CCT) and is measured in degrees Kelvin.

However, light sources can also be characterised by the quality of their light. Two light sources can have the same colour temperature but be very different in the composition of waves. One might be made up of a fairly even distribution of wavelengths, while the other might have a very uneven distribution with some wavelengths missing and others spiked up. Although these two sources will give the same overall colour cast to a scene the one with the very uneven spectral distribution is likely to strike particular 'coloured' objects in odd and unpredictable ways. Some colours may appear washed out, with others overemphasised. A light source's quality is rated using the Colour Rendering Index (CRI), which has a scale of 0-100. The closer to 100 a light source is, the more even the distribution of wavelengths and more 'natural' the light.

Substance

The second key element in our experience of colour is the substance upon which, or through which, light falls. Objects in the world are not really 'coloured'- they simply absorb, transmit or reflect particular wavelengths from visible light. If a collection of objects appear to us to be different colours, this is because each object differs in the way it responds under a light source.

A 'white' object reflects all or most of the light that falls upon it, while a 'black' object absorbs all or most of the light. Plants appears green because they have pigments which absorb wavelengths from the red and blue parts of the visible spectrum and only allow the 'green' wavelengths to be reflected on to a viewer.

This process of absorbing parts of the spectrum not only occurs when light is being reflected off solid surfaces, but also when it is transmitted through substances such as filters - or, indeed, the atmosphere. A bluish camera filter, for example, will absorb many of the longer green and red wavelengths approaching the camera and pass through more of the shorter 'bluer' wavelengths. Note that such a blue filter could be used to help correct the red-orange cast in the candlelit scene described above by suppressing the preponderance of red wavelengths within the scene.

In the same way that light sources can appear to produce light of the same colour, but actually be emitting very different combinations of light waves, substances can appear to us to be the same colour but actually be absorbing and reflecting light waves in quite different ways. This is an important point to grasp. As we'll see below, most of the "natural" colours we see in the world around us are capable of being simulated by different intensities of just three fairly narrow bands of the spectrum (either Red, Green and Blue, RGB, or Cyan, Magenta and Yellow, CMY).

Human Perception

The third key element in our experience of colour is the human visual system. Vision is a very personal, subjective experience, but the basic mechanics are the same for most of us. Light enters our eyes through the lens at the front of the eye and is focused onto the retina inside the back of the eye. The retina is covered with millions of light sensitive cells which pass on signals to the brain via the optic nerve.

Diagram 2. Sections through the human eye and the retina

There are two types of light sensitive cells in the retina, called rods and cones because of their shape. Each eye has about 120 million rods, which tend to be concentrated more around the outer edges of the retina. Rods are not sensitive to colour differences, but record information about lightness and darkness. They are good for detecting motion and for seeing in low light-levels.

The 6 million cones in each eyeball are sensitive to colour rather than lightness and are concentrated towards the centre of the retina where there is more light. Each cone contains photopigments which are attuned to a particular band of the spectrum. When they detect light within their range they produce an electrochemical response. In humans there are three main sorts of cone, which respond to either long, medium or short wavelengths. They're usually called red, green or blue cones because these are the predominant colours within each band. However, the spectral bands they detect are actually quite wide and overlap each other. Other animals have a smaller or greater number of different cones. Many animals also have cones capable of detecting ultraviolet light.

The way the information from the cones combines to provide our experience of colour is still not entirely understood, but it seems most likely that the outputs from all three cones are used to determine the colour and quality of the light by comparing the responses from the different cones.

As we'll see below, a 'trichromatic' model of colour perception was developed in the nineteenth century. This argued that the eye must have three different types of receptors for colour: each sensitive to red, green or blue; each combining to represent different colours. An alternative nineteenth-century theory, the 'opponent' model, argued that there were three very different sorts of receptors. These determined whether the colour was: black or white; red or green; or yellow or blue. Twentieth-century science has offered some support to both theories. Studies of the eye have found that there are indeed three different cones that react to 'red', 'green', and 'blue' wavelengths, while studies of human perception - such as our experience of contrasting colours and afterimages - and from colour deficiencies, like colour blindness, have suggested that there are red-green and blue-yellow oppositions involved in our perception of colour.

The most popular theory today, the 'opponent-process' theory, brings together both of the earlier theories. It suggests that the red-green-blue information recorded by the three cones is further processed within the retina to produce three different "channels" of information which are communicated, electrochemically, to the brain: red-green; yellow-blue; and white-black. According to this theory, information received from the red cones and green cones is compared to determine the blackness or whiteness of the light; information from the red and blue cones is compared with information from the green channel to determine its 'redness' or 'greenness'; and information from the blue cones is put against information from the green and red cones to determine its 'blueness' or 'yellowness'.

Don't worry about the detail - The key point to grasp is that our eyes are particularly sensitive to the red, green and blue wavelengths in the light and colours we see. Because of this, by presenting different intensities of red, green and blue light, it is possible to fool our eyes into thinking that they are seeing other colours. This principle underlies the practice of colour reproduction, enabling us to reproduce or simulate a full spectrum of colours from just three "primary" colours: red, green and blue, in 'additive' colour processes; or their complementaries, cyan, magenta and yellow, in 'subtractive' colour processes. We'll turn to these processes now.

Additive and Subtractive Colour

Colours can be created in one of two ways. Firstly, certain wavelengths can be subtracted from the full spectrum (by being absorbed by a substance) leaving the others to pass into our eyes where they are experienced as particular colours. This is known as the subtractive colour process. Most of our experience of colour in the world is due to the subtraction of wavelengths. As we said above, we experience a leaf as green because the leaf absorbs the other parts of the spectrum and only reflects lightwaves from the green part of the spectrum. Subtractive processes underlie reproductive techniques that rely on reflection, such as colour printing or film photography. Light hits the page or photograph, certain parts of the spectrum are absorbed and the remainder are reflected and experienced as colour.

An alternative approach to producing colour is to project lights that are already limited to particular bands of the spectrum (e.g. red and blue lights) and allow their lightwaves to combine to form other colours (in this example, magenta). Because this process adds lightwaves, it is known as the additive colour process. This is the way colour is reproduced on television and computer screens. If you were to put a magnifying glass to a white area of this monitor, you would see that it is actually composed of tiny dots of red green and blue light. These are small enough so that the light they omit seems to our eyes to be superimposed. When each of these three little coloured lights are at full strength, we will think we see white. When they are at varying strengths or intensities, our eyes and brains will interpret the lightwaves as other colours.

When reproducing colour, it is obviously more efficient and economical to use as few colours as possible. For centuries, philosophers, scientists and particularly artists have sought to identify the 'primary colours' that can be used to mix every other colour. Within the additive colour process, red, green and blue are the best choice for primaries, which is why lights of these colours are used within monitor displays. Red, green and blue work best because they directly match the way our eyes detect colour and light, with their red-, green- and blue-sensitive cones.

Digital images destined for the screen are usually held within an RGB colourspace (Red, Green and Blue), which is really just to say that each colour is described in terms of the strengths of red, green or blue required to reproduce it.

However in a subtractive colour process, like commercial printing, red green and blue are not the best choice of primary colours, because these colours subtract too much light. Imagine a blue dot on a page, with a red dot printed over the top of it. Light would hit the red dot and almost all of the green and blue lightwaves would be absorbed, leaving red waves to pass on. The blue dot would then subtract the red lightwaves and any of the remaining green, leaving little, if any, light at all to be reflected off the page. Within a subtractive system of colour reproduction Red, Green and Blue are incapable of producing the full range of colours we can see. The best choices are Cyan, Magenta and Yellow. This choice is not arbitrary: these colours have a special, 'complimentary,' relationship with red, green and blue.

Diagram 3. Subtractive colours (left) as if reflected from a printed page; Additive colours (right) as if projected by coloured lights. Illustrates the relationship between the additive primaries (Red, Green, Blue) and subtractive primaries (Cyan, Magenta, Yellow)

Here's how it works. Red, Green and Blue absorb most of the spectrum leaving only one band of wavelengths to transmit or reflect. Cyan, Magenta and Yellow, in contrast, subtract only one band of the spectrum (Red, Green or Blue, respectively) leaving the remaining wavelengths to reflect and mix together. The colour we identify as 'Cyan' is actually a mixture of lightwaves from the green and blue parts of the spectrum. A dot of cyan ink will subtract the red wavelengths from white light, leaving the green and blue wavelengths to reflect and combine to form the colour we call cyan.

By varying the strengths of cyan, magenta and yellow, it is possible to create all of the other colours we can see printed in books or magazines. If we wanted to create red, for example, we would overlay a dot of magenta and a dot of yellow. The magenta would absorb green and reflect both red and blue light. The yellow would then absorb the blue, allowing just the red to reflect into our eyes. In theory, we could add a dot of cyan on top of these and this would absorb the red, allowing no light at all to escape (i.e. black). In practice, because inks are seldom pure, a little light would be reflected and the black wouldn't appear completely black. For this reason, printers usually use a fourth colour, black, to make sure their images appear as dark as they should. In addition to improving the quality, black dots are also often mixed in as a cost-saving measure, since black ink is much cheaper than coloured ink.

Digital images intended for commercial offset printing are typically held within a CMYK colour space (Cyan, Magenta, Yellow and the "Key" colour, black). This is simply to say that each colour is described in terms of the strengths of Cyan, Magenta, Yellow or Black required to produce that colour image. We'll say some more about colour spaces later in this document and in JISC Digital Media's second colour document: Colour Management in Practice.

Links to further reading

• Breaking the code of color
  Explanation of colour vision
• Color Deficient Vision
  Useful information and resources on colour blindness
• Color FAQ / Charles Poynton
  Popular, but quite mathematical, FAQ on various aspects of colour
• Color Reference Library / Earl F. Glynn
  Very extensive links to all aspects of colour
• Color Matters / Jill Morgan
  A lot of information about colour science, aesthetics and psychology from an academic and professional colour consultant
• Colour Vision / Bruce McEvoy
  A wealth of information about colour vision and colour theory, covering art, psychology and science
• Colour Vision FAQ / Rochester Institute of Technology
  Good overview of colour vision
• FAQ about color physics / Stephen Westland
  Covers various aspects of the physics of colour
• Glossary of Color Science / Alex Byrne and David Hilbert
  A one-page glossary of terms used in colour science
• Joy of Visual Perception / Peter Kaiser
  An extensive online book on human visual perception
• Physics2000
  Excellent interactive tutorials on various aspects of phsyics, including:
  
  • TV Screens
  • Laptop Screens
• WebExhibits
  Series of well presented online exhibits on issues of art and science, including two specifically related to colour theory:
  
  • Color Vision and Art
  • Causes of Color: why are things colored?

Modelling Colour

Colour management relies on models or representations of colour. Over the centuries, many different models have been devised, based on different understandings or theories of colour and on different purposes, such as describing, matching, displaying or reproducing colour.

We've already encountered two colour models which are primarily used in the reproduction of colour: RGB, used in additive colour reproduction, such as the computer screen; and CMY (or CMYK), which forms the basis of subtractive colour processes, like printing.

This section presents a potted history of colour modelling, ending with the work of the CIE (Commission International de l'Eclairage, International Commission on Illumination) and ICC (International Colour Consortium), both of which have been very influential in producing models that enable accurate colour management within a digital context. Their work forms the basis of colour profiles which are described in more detail in JISC Digital Media's second colour document: Colour Management in Practice.

Historical Models

Colour theory has a long history: the earliest records we have of colour modelling date from ancient Greece. Aristotle, for example, felt there were seven significant steps of colour stretching from white through to black (his intermediates varied, but often included yellow, red, violet, green, and blue). He derived his hues from observations of the changing colour of daylight, but failed to understand that colour resulted from the properties of light itself. He believed instead that colours resided on the surface of objects.

Aristotle's medieval translator, Robert Grosseteste, introduced a second dimension in his model of colour - that of "lightness". He felt that Aristotle had been wrong to place white and black alongside the other colours, suggesting that these formed a second dimension, with the seven (unidentified) hues shading towards white or black.

In the Renaissance, artists like Leon Battista Alberti and Leonado da Vinci took a more practical approach to colour - seeking to identify the primary pigments they could use to mix all other colours. Both identified yellow, green, blue, and red as primaries, though Leonardo added white and black. Battista preferred to approach it like Grosseteste, treating white and black as a separate dimension to colour. Battista's 'model' can be represented visually as a double pyramid, with the four 'colours' occupying the four corners of the base and shading up towards white or down towards black.

In the following centuries, these kind of approaches evolved into complex colour wheels, pyramids or spheres, almost invariably based on the mixing of paint pigments. In theoretical explanations, colours were often described as mixtures of the four elements - fire, air, water and earth - along with 'lightness' and 'darkness'.

A significant breakthrough was made by Isaac Newton at the end of the seventeenth century. Others had used prisms before him to produce spectrums, but these had been interpreted as changing or colouring white light. Newton argued (and proved) that the prism was not really changing the light, but was splitting it up into its component parts. He produced a simple colour wheel with the seven colours of the rainbow stretched around its rim and a white hole in the middle, indicating that when all these colours were mixed together they would form white. While this was an important discovery, we should note that Newton's model was as much aesthetic as scientific: he chose seven colours because a musical octave has seven sound intervals; and in bending the straight spectrum to form a neat circle he was actually joining together two colour bands (red and violet) that are not naturally adjacent in the spectrum.

Newton's colour theory was opposed by the influential philosopher Johanes Wolfgang Goethe in the early 19th century. Goethe produced a triangle diagram with blue, red and yellow at its vertices. Within this he placed smaller triangles of secondary and tertiary colours. Goethe was not challenging Newton on scientific grounds, but on artistic or experiential grounds. He described and arranged his colours according to the 'moods' they could express or evoke. In doing so he was looked backward to the artist's practice of colour mixing and was anticipating the modern discipline of colour psychology.

For our purposes, a more important trajectory from the late eighteenth century was the development of the idea that colour could and should be represented with a third dimension. Colour had often been represented on 3-D spheres, cones and pyramids, but the colours were always wrapped around the shapes. For example, a globe might have pure hues of colour painted around its equator shading towards a white or black North or South pole, but no one had really considered that colour might exist within the shape, with the colours becoming less and less 'pure' as they approach a neutral grey core. Johann Heinrich Lambert proposed something like this using a pyramid shape in the 1770s, and Philippe Otto Runge produced a solid sphere of colour in the 1810s - around the same time as Goethe was producing his triangles. Runge actually adopted Goethe's red-blue-yellow primaries as the basis for his colour model. In the early twentieth century this three dimensional approach to modelling colour found an expression in the physical models constructed by the art teacher Albert Munsell. It also formed the basis for later perceptual colour models like HSL - Hue, Saturation and Luminance - a system used in some computer colour modelling.

By the nineteenth century the distinction between subtractive and additive colour had become clearer and important progress was made on understanding additive colour mixing. At the beginning of that century, Thomas Young proposed that all the colours we see could be produced from three wavelengths (red, green and blue) and that the human eye had three receptors to detect these (the 'trichromatic' theory). Young's ideas were more fully developed and systematically investigated by the physicist James Clerk Maxwell mid-century. Maxwell adopted the red, green, blue model and developed a method and a notation for measuring and describing additive colour. He developed his model by spinning disks painted with different percentages of colour and asking people to describe and match what they saw. His approach formed the basis of the science of colorimetry and was developed further in the twentieth century by the CIE, ultimately resulting in the ICC profiles, which we describe below.

Contemporary Models

By the early twentieth century, the nature of colour was fairly well understood: the distinction and relationship between additive and subtractive colour was clear, as was the understanding that colours could be described in terms of their hue (particular colour), strength (saturation or intensity), and brightness (lightness or darkness).

There was an increasing need in the twentieth century to develop colour reference systems that would enable people to talk about colour and know that they were referring to exactly the same thing. Many such systems were devised, but two that are well known are the Munsell system and the Pantone system.

The Munsell Color Order System grew out of the work of art lecturer Albert Munsell, in the early years of the twentieth century. He felt that earlier attempts to model and describe colour hadn't adequately represented the relationships between colours as humans perceive them. Like others, he saw a three-dimensional model of colour (hue, strength and brightness) as the most useful, but he felt that forcing the colours into a symmetrical shape was wrong. He also strongly objected to placing the 'pure' colours around the equator of a globe because, in his judgment, yellow was brighter than red and red was brighter than blue. Using coloured spinning tops (like Maxwell and others) and relying on his own visual sense, Munsell built a three-dimensional model in which he felt the steps between the adjacent colours were exactly equal. For some hues this resulted in more steps than others, producing a quite irregular shape. Because the result was not a sphere, he called it a colour tree.

The Munsell Color Tree (image courtesy of Michael Sitko)

Munsell devised a notation for referring to the colours on his colour tree and published his system in books and charts. As a widely available reference with a systematic form of notation, Munsell's system enabled people who were using colour to be very specific about which colours they were referring to. The Munsell Color Ordering System was endorsed by a number of standards bodies and is still in use today.

The Pantone Color system, developed in the 1960s, is also intended to enable people to refer precisely to colour, where such precision is critical (e.g. fashion and manufacturing industries). Unlike Munsell and many other colour reference systems, Pantone colours are specificed as full strength (i.e. fully saturated) hues of colour, although in practice they are used at different percentages to produce varying tints of the colour There are many different sets of Pantone colours, designed for different applications (e.g. paint or dye mixing, printing on matt or glossy paper). Coloured chips, palettes - and now computer software - is used to aid in matching Pantone colours for different applications.

Reference systems like Munsell and Pantone rely, largely, on the human eye to make colour comparisons and matches. The same is true of traditional printing processes - humans were used to check the quality of the colour reproduction from a particular print set-up and make any compensations or adjustments that were necessary. But as computers became more and more involved in colour display and reproduction, the need to guarantee colour consistency became increasingly important.

As we've said earlier, digital image files could be encoded as RGB values for computer display or CMYK values for some printing purposes. However, since the quality of the lights (phosphors) on a monitor can vary and printing inks can behave very differently, the same set of RGB values can look very different on different screens and a single CMYK file can print differently according to the printer used. There are further challenges in converting an RGB file for the screen to a CMYK file for printing, since the colours are being mapped from an additive colour system to a subtractive colour system each of which will convey the colour in a fundamentally different way.

RGB and CMYK are 'device-dependent' colour spaces which means that the colour of the described value will be dependant upon the capability of the device. What is needed is some 'device-independent' system that will describe the colours independently from the device so that we can match our colours across a range of different devices. This is the challenge Colour Management seeks to meet and we turn now to its development.

CIE

We said earlier that seemingly identical light sources or colours can actually be composed of very different distributions of wavelengths. This is the whole basis of additive and subtractive colour: using a small number of colours to simulate every other colour. As we suggested, the reason we are so easily fooled by this is that the human visual system (eye plus brain) is doing its colour interpretation based on just three sensors (red, green, blue).

But in order to accurately control and reproduce colour within an additive (or subtractive) colour system, we must know exactly what combination of primaries is required to match and simulate each real-world colour. This is the challenge that science of colorimetry has set itself: to determine and map out these combinations. Each combination of red, green and blue are known as the tristimulus values for the corresponding real-world colour.

The characteristics of the original, real-world colour can be measured objectively using instruments to detect the wavelengths they reflect (the results are referred to as the spectral power distribution for that colour). But since the tristimulus (red, green, blue) equivalents depend on human perception, they must be measured using human observers rather than instruments.

In 1931 the CIE undertook the huge task of matching real-world colours with their tristimulus equivalents. In an approach reminiscent of Maxwell and Munsell with their spinning tops, they put each test colour alongside another colour created by mixing red, green and blue lights. The strengths of the red, green and blue lights were adjusted until the human observer considered the two colours were a perfect match. Lots of human observers were involved and the results were averaged and published by the CIE as the 'Standard Observer'. In 1931 the observers were viewing the colours through small slits which gave them a 2 percent field of vision. This was later judged too limiting and the experiment was repeated in 1956 with 10 percent vision allowed and a wider selection of subjects with a more diverse ethnic background. Many of the mathematical or computer colour models we use today are based on the data from these two experiments.

CIEXYZ colour model

The results of the 1931 tests were laid out on a graph in a horseshoe shape (see the illustration below). This shape results from a particular mathematical equation used to relate the data and it could have been drawn very differently. What is significant is that it represents every hue - and every saturation of that hue - that a human being is capable of seeing. What it doesn't show are lightness differences. These must be imagined as a third dimension coming out from the page. Nor does it represent accurately the differences or distances (i.e. the size of the steps) that humans perceive between the individual colours.

Because it lays out the colours in two dimensions, the CIE horseshoe is useful for visualising and comparing colour 'gamuts'. We said above that RGB and CMYK colour spaces are 'device-dependent'. The 'gamut' is the range of colours capable of being produced by a particular device. (e.g. printer or monitor). As the diagram below shows, colour gamuts are generally more limited than the human eye.

Diagram 4. CIEXYZ model, with several colour gamuts mapped

In this diagram, the white shape represents the colour captured by a particular scanner; the green, the colours available on one monitor; and the black the colours capable of being printed by one particular printer. This demonstrates the whole reason why colour management is so important - if not carefully managed it is so easy to lose or misrepresent colour as image files are exchanged between particular devices.

CIELAB colour model

The 1931 CIEXYZ model was important and influential, but by no means perfect -especially in the way it failed to represent accurately the perceptual distances between colours. Many attempts were made to improve on it throughout the twentieth century. One of the most successful - or more influential - was the CIE's L*a*b* model (or CIELAB), developed in 1976. CIELAB still relies on the 'Standard Observer' data collected by CIE, but it uses another set of mathematical equations to plot the data in a very different way.

Unlike the CIEXYZ model, CIELAB is three-dimensional. It also accurately represents the steps between colours. In CIELAB, one axis (a*) plots values between red and green; another between blue and yellow (b*); while the third axis plots the lightness or luminance (L*) from white to black. This is difficult to represent on the page, but the drawing below is an attempt.

Diagram 5. Model of CIELAB

Individual colours are referenced according to their positions on all three axes. In a sense, CIELAB is very reminiscent of the three-dimensional models of the past, especially Munsell's model. Like Munsell, the steps between points within the CIELAB shape or colour space are intended to be perceptually accurate - to reflect the degree of change between colours, or saturations, or brightness levels as a human being perceives them.

Like CIEXYZ, CIELAB offers a reference colour space within which particular colour gamuts can be compared. It can also be used as an independent, intermediary colour space: values from a particular gamut can be re-encoded as CIELAB values and then other devices can take those values and covert them into their own colour gamut. CIEXYZ could and has been used in this way, but CIELAB is a mathematically or computationally superior solution.

ICC

The ICC> (International Colour Consortium) was formed in 1993 by several large industry players (including Adobe, Apple, Microsoft and Kodak) and was based on the original work of the Colorsync consortium that was supported by Apple Macintosh. Its purpose was to develop a universal system for managing digital colour. ICC recognised that the best approach was to use an intermediary device-independent colour space as a means of translating colour from one gamut to another, as we've just described. They chose CIEXYZ and CIELAB as their independent colour reference, and designated the environment within which the translation takes place as the 'Profile Connection Space' (PCS). 'Profiles' are files which describe the particular characteristics (or biases) of a particular device. They enable a colour management system to accurately map colour values between device-dependent gamuts and the device-independent reference colour space. Apple had already done a lot of work on colour management, so their ColorSync profile format was chosen as the basis for the ICC profile format.

This has been a brief introduction to ICC. JISC Digital Media's second colour management paper, Colour Management in Practice, describes the practicalities of using ICC profiles within a colour-managed workflow.

Other approaches

Most acknowledge that ICC profiles are the best way to define and manage digital colour. However, not everyone is going to go to the trouble of calibrating and profiling their monitors and other devices. In addition, a lot of computing power is required to make the necessary colour transformations required within a properly ICC-managed environment. Because of this Microsoft and Hewlett Packard sponsored the development of sRGB, or Standard RGB.

sRGB is based on the RGB values of an average PC monitor. Images encoded in sRGB are in a known colour space, which can be used by devices (or software) that understand that colour space. Using sRGB is much better than leaving it all to chance, but is not as good as using ICC profiles and large colour spaces. This is because sRGB is a very small colour space, representing the limitations of computer monitors (see the CIEXYZ diagram above). CIEXYZ or CIELAB, in comparison, are very large, reflecting the full potential of human vision. Typically, an image capture device (e.g. scanner or camera) is somewhere in between these. What this means is that if the colour information received from the capture device were transformed into CIELAB, all of the colour information would be retained. But if it were transformed into sRGB, some of the colours would be lost as the colour information was compressed or clipped.

Another alternative is Adobe RGB 1998. This colourspace was developed by Adobe and it is now on track to become an international standard. Adobe RGB 1998 is very much bigger than sRGB, so raw image data transformed into Adobe RGB 1998 does not suffer the loss that it would if transformed into sRGB. However, while Adobe RGB 1998 is a superior colour space to work in if you're saving or optimising digital images, it is not an ideal space to use when delivering images to a computer screen. This is because monitors have more or less the same limitations as sRGB (obvious, when you remember that sRGB is based on average monitor capabilities!). If you deliver an Adobe RGB 1998 image to a screen, you are gambling on how the monitor will transform the colour information. In an ideal world, an Adobe RGB 1998 image would be individually transformed into an image that matches the gamut of the monitor on which it is to be viewed. While it is entirely possible to do this for your own monitor (using ICC profiles), this is impossible if you're letting the image loose on the World Wide Web. In that environment, the best you can currently do is transform it into sRGB.

Links to further reading

• Bruce Lindbloom.com
  Lindbloom's site contains some useful tools for modelling and calculating colour
• Bruce Fraser's CreativePro page
  Fraser is a prolific writer on colour management issues. This page provides links to his colour articles on the CreativePro.com Web site
• Color Models / Colorcube
  Includes graphical illustrations of various historical colour models
• Color Systems / Hans Irtel
  Reproductions of historical colour models
• Colour Museum: Colour Order systems in art and science / Narciso Silvestrini and Ernst Peter Fischer
  Presents nearly 60 different historical colour theories
• X-Rite/GretagMacbeth
  Owners of the Munsell colour system
• International Color Consortium (ICC)
  Official home of the ICC profiles.
• Pantone
  Official Web site of the Pantone colour system

Last updated: 16 December 2008
Published in: Digitising analogue media | Creating new digital media |
Tags: colour management | workflow |

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Comments (1)