Colours, computers and colour management

Colour is something that most of us take for granted, but for manufacturers, lack of colour control can be a serious economic problem. Visible inconsistency in paints, inks and dyes costs money. For example, a car manufacturer couldn’t sell cars if a particular paint colour varied from one car to the next. For product advertising, manufacturers need be able accurately to represent a product’s colour in print, on the Internet, in film and on television. For this reason, as soon as it was possible for industries to produce a wide range of colour pigments, our perception of colour, and the measurement and control of colour, became the subject of a great deal of study.

At home and at work even computer users who aren’t colour professionals are finding that they need to manipulate colour images. Intentional colour management provides the means both to control and reproduce colours as accurately as possible.

To apply colour management effectively you need to know the colour theory behind it. This article introduces some of the ideas of colour theory and shows how it is used in digital colour management.

A Brief History Of Colour
Humans have been using coloured pigments for thousands of years, although for much of this time only a limited range of colours were available. The use of oils blended with new pigments by Jan van Eyck, best known for his 1434 painting of “Giovanni Arnolfini and his wife”, revolutionised painting. However it is only since the eighteenth century that a really wide range of low cost colour pigments have been available. The Industrial Revolution saw the invention of aniline dies from coal products in the 1800s and the first colour photograph was made in 1861.
The earliest colour space, still in use, the CIE XYZ space, wasn’t published until 1931 (colour spaces will be described later). Apple incorporated basic colour management into its operating system quite early on and was instrumental in establishing the International Color Consortium, or ICC (www.color.org) in 1993. Apple’s early support of colour control helped to establish its products in the field of art and design, and area in which the company has had considerable success. Microsoft Windows did not provide any colour management facility until Windows 98, and colour management tools have only recently appeared in Linux.
One of the problems in understanding colour management is that there are numerous standards and methodologies, each a product of the time it was introduced and the particular colour problems it was intended to address. For example, instead of just one colour model or space there are several, which can be confusing to the colour management novice. Another problem is that colour and colour management is a complicated subject and a great deal of the available information is either misleading or just plain wrong.

Colour Does Not Exist
Strictly speaking colour is a product of the human visual system. All that can physically be measured are the wavelength and intensity of light and not its colour. Certain mixtures of wavelengths can only be referred to as purple or brown because of a general agreement that that is what we call the visual sensations they produce. Since colour is a subjective sensation, colours are not experienced in the same way by everyone. You might think this rather difficult to prove, since we cannot directly experience the thoughts and sensations of others, but about two per cent of females and eight per cent of males have some degree of colour blindness. The most common form of colour blindness is rather misleadingly referred to as red-green colour blindness.

A great deal of the effort in colour science has gone in to methods of relating physical measurement to the human perception of colour. Unfortunately this is where most of the confusion over colour control and management arises. For one thing our visual perception system is not linear, so colour systems based directly on linear measurements (like the CIE XYZ 1931) do not map well to our experience of colour. For example there is quite a large portion of the green area of the CIE XYZ space that is perceived by humans as being all the same colour.

Comparing Senses – Hearing To Sight
There have been attempts to draw parallels between music and colour and to produce a system to organise colours, in the same way that musical theory provides structures such as scales and harmony for musical composition. However the human auditory and visual systems are rather different. The average human can hear a range of over 9 octaves and the common musical instrument with the widest range, the concert grand piano has a range of 8 octaves. Most human voices can cover 2 to 3 octaves. Each musical octave is a doubling of frequency. If you apply the same idea to the range of frequency of visible light you will see that there is a span of slightly less than one “octave”. Visible wavelengths range from the short 380 nm (nanometre) for violet to 750 nm for the long wavelength red. Generally we speak of only seven pure colours in the visible spectrum – violet, indigo, blue, green, yellow orange and red. However there has been some success with ideas of colour harmony and various colour wheels are used by artists and designers to choose colours that go well together. Colours used for a particular design project are often organised in a palette. Colour palettes are commonly found in design software such as CorelDRAW.

Unlike our hearing, which can sense and identify sounds of a single wavelength, our eyes have three sets of sensors, known as cone cells, each of which senses light over a range of frequencies. The maximum sensitivity of one of these sets of sensors peaks in the blue, one in the green and one in the red. Outside their peak the sensitivity of each group of sensors falls off smoothly and the green sensors overlap with the blue and the red. Colours that fall outside the peak sensitivity of any of the sensors are recognised because of the balance of sensations they produce between two sensors. For example yellow stimulates both the green and the red sensors. Part of the reason our colour vision works so apparently well with just three sensors is because of the fairly limited human visual range. Our eyes also have rod cells, which respond only to light intensity.
Colour displays and printers rely on the way our eyes work and create a wide range of colours by mixing only three colours together.

Emitted, Reflected – Additive, Subtractive
Humans experience colour in two ways: by emitted light directly entering the eye or, by light reflecting from the objects around us. Mostly we experience colour through reflection and this means that the appearance of colour depends on the the nature of the surface of the illuminated object and on the spectral content of the light that is illuminating it. For example a sheet of “white” paper appears red when red light is shone on it and green when illuminated with green light. A sheet of red paper illuminated with green light appears black. This means that when working with colour and colour management the nature of the light that was and/or will be used as the illuminant must be taken into account.

Primary And Secondary - RGB Or CMY / CMYK
All the colour imaging devices we use that emit light – computer monitors, televisions and projectors use a mixture of red, green and blue light to produce a wide range of perceived colours. These devices use what’s called additive colour mixing. Red, green and blue are referred to as primary colours because they relate directly to the three groups of sensors in our eyes. On the other hand, printers use cyan, magenta and yellow and so-called subtractive colour mixing. These three colours are referred to as secondary or complementary colours because cyan is white minus all the red frequencies, magenta is white minus all the green frequencies and yellow is white minus all the Blue frequencies. Or to put it another way, the three secondaries can be obtained by mixing pairs of primaries as follows; mixing blue and green light produces cyan, mixing red and blue gives you magenta and mixing red and green makes yellow.

Why Are There Two Methods Of Producing Colour?
The reason that we use two methods of producing colour is this: for displays, the base state is black, so emitted light has to be added to that base state to produce colours; for printers the base state is the “white” of the paper and colours have to be selectively subtracted from that base state, leaving the required colour reflected from the page. This is why the paper used has such an effect on the final image quality and why using the more expensive, “whiter” papers results in a better image (unlike many paints, most printer inks are transparent to some degree). The expensive photo papers also have smoother surfaces and may have a gloss surface to reduce diffusion of the light striking the surface, resulting in greater image contrast. The variation in paper “white-ness” and the other variations in the nature of paper are the reason why printer drivers and colour managed applications have a choice in their output settings for paper type.
The pigments and dyes used for printing inks and the process of printing on paper aren’t as successful as the three colour system used for displays. Printing cyan, magenta and yellow inks to get black usually results in a dark brown, so all ink colour printing uses an additional pigment – black - to form the CMYK system of colour printing (the letter K from the end of the word black is used to avoid confusion with blue). Photo printers extend the range of printable colours still further by adding more ink colours. Usually these are lighter versions of cyan, magenta and yellow.

White And Colour Temperature
In abstract, pure white might be regarded as emitted or reflected light that contains equal amplitudes of all the visible frequencies. Such pure white light never occurs in nature or in colour management. Our main source of illumination, sunlight, does not have a smooth or equal energy distribution and its frequency content varies depending on the time of day and on weather conditions. Artificial light is even less “pure”; for example, fluorescent light has a very uneven energy distribution with a number of large amplitude spikes at certain frequencies.
In practice white is always variable and relative and the human visual system is extremely good at adjusting, so that, under a wide range of lighting conditions, the areas of view that emit or reflect the widest range of frequencies with similar strength are usually seen as “white”. Colour film cameras, lacking any built in compensation, don’t do this and the best that can be done is to use film optimised for different lighting conditions such as “daylight” or “tungsten” film and/or to supplement these with coloured filters fitted to the lens. Perhaps one of the biggest advantages of digital cameras is that they have either, or both, auto and manual white balance. Even if the colour balance out of the camera does not look right it’s easily tweaked with photo editing software, which frequently has an auto-colour or auto-white balance control.

The variations in what may be considered white mean that this has to be taken account of in colour management. Colour temperature is often used as a reference for white. It’s based on the idea that a perfect black body radiates light according to its temperature. At zero Kelvin a black body radiates nothing, at 5,000 K it radiates yellow-white light rather like morning daylight, at 6,500 K it radiates blueish-white similar to overcast daylight at noon and at 9,300 K it radiates a hard blue-white light. D50 and D65 (for Daylight 5000 and 6500) are commonly used as lighting references. 9,300 K was frequently used as a default monitor setting because monitors were most efficient at this setting and produced high brightness.

Representing Colour Using Colour Spaces
One of the most confusing things you may ever see in a text book on colour is the 2D representation of the 1931 CIE XYZ colour chart for RGB displays. It is rarely explained that this colour chart represents a plan view or projection from above the white point of a 3D volume and is shown this way purely for convenience, given the difficulty of representing 3D objects in print. All colour spaces are three dimensional volumes. The vertical axis of these spaces always represents luminance or brightness and runs from black at the bottom, through various shades of grey, to white at the top.
The most commonly used reference space today is the 1976 CIE L*a*b* (pronounced as elstar ehstar, beestar). Rather confusingly this is often referred to as Lab but there is an earlier space, the 1948 Hunter Lab space, to which that nomenclature more properly belongs. L*, standing for Luminance is the vertical axis and the orthogonal axes a* and b* are red-to-green and blue-to-yellow. L*a*b* is often used as a reference colour space for performing transformations from one colour space to another because, at present, it is the space that comes close to perceptual linearity. For example the popular Adobe applications such as Adobe Photoshop use a version of L*a*b* as their internal colour model.

Saturation, Hues, Tints And Shades
A special vocabulary is used for describing colour. Unfortunately these terms are often misunderstood or misused and in some cases are hard to define. They are saturation, hue, shade and tint.

The strongest, purest colour is said to be saturated. For displayed digital images fully saturated colours are represented by the largest numbers, or 255 for a 24-bit image (32-bit colour usually uses the extra 8 bits to represent other image attributes such as transparency). Very pure saturated colours are quite rare in nature because most naturally occurring colours are mixtures of more than a single wavelength. Artificially produced, single wavelength colours, are fully saturated.

Colour hue is the attribute that describes what appears to be the colours dominant wavelength. For example the hue for pink would be red.
For mixing opaque pigments, as in paint, tints are achieved by mixing a colour with white, while shades are achieved by mixing black with a colour. For example light pink is red mixed with white, while dark pink is black mixed with red. For colour produced by emitted light, such as from a display screen it is always a question of mixing the colour with white because the base state of all displays is black. For displays, the shade attribute of a colour can be regarded as brightness or luminance.

Colour Gamuts
Every colour device has a gamut, a range of colours it can capture or reproduce. This gamut can be represented as a device dependant colour space within the framework of any of the reference colour spaces. The term gamut is used to refer to the relative volume of the colour space. It is often also used to refer to the area of the 2D projection of a colour space, since this is related to the volume and represents the range of saturated colours. The human visual system - for most people - has a large gamut. Frequently this is referred to as containing 65 million colours. The human visual system is not good at recognising absolute colours, but is much better at recognising the difference between discrete patches of similar shades. The figure of 65 million recognisable shades has been determined by experiment as the average colour discrimination for most people.

Colour Intent
Colour intents are one of the most difficult aspects of colour management to understand. Users of colour management systems are required to choose which intent to use. Intents are needed because none of the current elements in a colour capture, manipulation and reproduction chain can match the gamut of the human visual system. All capture devices, displays and printers have a smaller gamut. Colour printers usually have the smallest gamut, or device colour space, of all the colour devices we use. Some of the colours we can see cannot be captured, and some of the colours that a camera or scanner can capture, cannot be reproduced. This leaves us with the problem with what to do with the colours we have captured in an image but cannot reproduce in print and here is where the choice of intent comes in.

A colour intent describes how any colours that fall outside the gamut of a reproduction device, normally a printer, are rendered. There are four intents in common use; perceptual, relative colorimetric, absolute colorimetric and saturation.

Perceptual rendering compresses the source gamut so all the out-of-gamut colours fit inside the destination gamut. This can result in some distortion of the relationships between colours. With perceptual rendering none of the original colour information is lost and in theory the transformation could be reversed in order to restore the original image.

Relative colorimetric rendering maintains a near exact relationship between in-gamut colours but may clip colours that are out of gamut. Clipped colours are lost and cannot be recovered.

Perceptual or relative colorimetric intents are the best choices for photo realistic images.

The absolute colorimetric intent is similar to relative, in that it clips out-of-gamut colours and preserves those that are in gamut, but it treats the white point differently. With absolute the white point does not change, while with relative it may, if this is required to maintain the relationship between colours. This means that with a relative conversion the appearance of whites in an image may change, they may get slightly more red or blue, while with absolute they won’t.
Absolute rendering is used when it is important that the colours that are reproduced remain as accurate as possible.

Saturation rendering tries to preserve the saturation of colours while making no attempt to reproduce photo realistic images. It is a good choice for business graphics such as charts.

The Colour Management Workflow
A colour management system consists of a number of hardware devices, each with their own gamut or colour space, with an image in the form of digital data being passed from one device to another. As the data moves between each device it often needs to be translated from one colour space to another. To do this it is necessary to know the source and destination colour spaces used, the devices must be calibrated and a profile that describes the devices’ colour gamut must be used.

A typical colour system chain consists of a camera or scanner passing data to a computer running a colour management application. Edited colour images from the colour management application are output to a web page, a local colour printer or sent to a commercial colour press. A calibrated and profiled colour display is used to view the colour images at various stages in the process.
A properly calibrated and profiled colour display is absolutely vital to establishing colour control because it is the user’s window on the entire process. These must be calibrated and profiled using a hardware monitor calibrator such as the ColorVision Spyder 2. Older displays that have been in use for several years are often unsuitable because they can no longer generate the peak brightness that is required for accurate calibration. Although LCD monitors are now the usual display type many of them are unsuitable for serious colour work, either because they lack the necessary user controls for calibration, or are limited to only 18-bit colour rather than 24-bit.

Colour And Computers
The majority of computer users expect to be able to use a digital camera or a scanner to capture an external scene, to edit and manipulate those images and then to print or display them on-screen without any unexpected changes in colour balance. Most people would probably say that they would expect a reproduced image on screen or printed on paper to “look the same” as the original scene. It would be very bad for business if, when operated by the average user, computers displayed or printed colour images where the colour balance was obviously wrong. In the early days of colour computing this is frequently exactly what did happen.

This problem has been solved by building automatic colour management into the operating system and into printer drivers. It works by using a colour space that has a relatively small gamut for reference and by assuming that, by default, everything in the colour work flow conforms to that colour space. The space used, sRGB (standard Red Green Blue) was originally adopted by Microsoft and HP for standardising colour on web pages and is based on the colour space of the typical computer monitor (at the time, this was CRT). Therefore most monitors conform reasonably well to sRGB without any calibration or profile. Many digital cameras, especially the consumer models, use sRGB as their default colour space, although it may be possible to select other colour spaces.
Instead of individual colour profiles measured from the actual devices being used, manufacturers’ generic profiles are used. Generic profiles, which allow for a certain amount of manufacturing variation, are usually included with most colour devices, or are available from websites.

When it comes to printing, all printer manufacturers have a selection of presets included in their printer drivers designed for a range of papers, although these are of course always designed for use with the printer manufacturer’s own brand of papers. These profiles are often supplemented by settings for things like “best quality photo” or “text and image” which are effectively simple ways of setting the colour intent.

Although automatic sRGB works reasonably well and is a lot cheaper than a fully colour managed system, intentional colour management allows more accurate control of colour and can provide better results.

The Colour Management Chain – Open Or Closed Loop
A colour management work flow can be regarded as a closed loop when the characteristics of the final reproduction device can be measured and fed back into the work flow. If the final colour output is to be a commercial printing press the ideal would be to obtain an individual device profile for the press, inks and paper used and to use that to create the output file. With such a profile it is possible to close the management loop. Commercial print press operators in general have always proved to be a huge stumbling block in this respect. It is normally extremely difficult to establish a meaningful dialogue with commercial printers on the subject of colour in general or on press profiling.

Most printers will supply hard copy colour proofs on request and for a fee. The problem with these is that usually they are not produced using the same press characteristics, inks and paper used for the final print run, so they are only useful for showing really gross errors.

What is possible is to establish to which standards a commercial printer says they calibrate their presses, and what type of paper they will use. This is likely to be one of the commercial standard profiles which are supplied with leading image editing software. Adobe Photoshop for example has profiles for SWOP (Standard Web Offset Press), Euroscale coated V2 and so on.

In Conclusion
In essence, colour management simply consist in choosing and applying the correct colour spaces and providing accurate device profiles at every step of the colour management work flow. In the past, poor software implementation, lack of accurate low cost hardware and software tools and poor support for colour management within the operating system, have made it hard for the average person to establish a colour management work flow that actually works.
Fortunately over the years the situation has improved and software and hardware colour management systems that leverage the latest advances in technology, such as high intensity LED’s, instead of exotic gas lamps, are becoming available. It is now possible to set up a calibrated and profiled colour work flow on an ordinary desktop PC at reasonable cost.