AGOCG logo
Graphics Multimedia VR Visualisation Contents
Training Reports Workshops Briefings Index
This report is also available as an Acrobat file.
Back Next Contents

2.2 CMS Principles

The problem of colour fidelity is due to the different methods devices have for creating colour together with inter-device variability. Monitors work in RGB colour spaces where the colours are created by mixing proportions of Red, Green and Blue light. However, monitors from different suppliers may use different phosphors and an individual monitor itself will age. This is equivalent to different or gradually changing colour spaces. Consequently specifications in RGB do not necessarily correspond to a unique colour, even on the same monitor.

Printers, generally, do not work in terms of the addition of light, but in terms of the addition of pigments. This is equivalent to the subtraction of coloured light from an incident, usually white, light. They work in proportions of three primaries, Cyan, Magenta, Yellow and Black, or CMYK colour space. As in monitors, the effects of fixed proportions of CMYK will vary depending on the nature of an individual printer's basic inks and the paper used.

Further problems are introduced by the practice of using scanners to read in images, as scanners work in yet another colour space, YCC. Similarly, other classes of input and output device have their own colour spaces.

Colour Management systems therefore have to convert between different colour spaces. The basic components of a CMS are some means of representing colour information in device independent terms, some means of converting between that representation and the colour spaces of the various devices, and some means of matching input colour to output colour when no exact match is available.

The basis of the device independent colour representation is usually the CIE XYZ space. This is based on the standard CIE colour observer without reference to any particular device. Colour Management Systems contain device profiles that essentially allow conversion between the device's own colour space and the underlying common reference colour space. Profiles are constructed by various mixtures of calibration software, spectrophotometers or standard targets of known specification (such as ANSI IT8), depending on the nature of the peripheral device.

For a closed system, with known input and output devices, this process of colour space conversion can be customised and this is the approach which has been adopted for critical applications in the recent past.

Even with a closed system problems may arise. A device using one colour space can produce colours that cannot be represented on a device using a different colour space. The range of colours each device can produce is termed its gamut. Generally speaking the gamut of additive devices, e.g. monitors, exceeds that of subtractive devices, e.g. printers. Consequently it is not difficult to generate colours on a monitor not achievable on a printer. When an input colour lies outside the gamut of the output device, the CMS has to provide some means of obtaining the best possible match.

In these cases the CMS has to be able to work out some best fit equivalent. Packages incorporating CMS often access the CMS when the user is creating a colour palette to warn the user if an out-of-gamut colour is being created. This allows the user to alter the colour choice if necessary. If the user decides to keep the colour, or the colour exists in an imported image and the user is not able to alter it, then colour matching algorithms come into use.

Each Colour Management System will have its own approach to this problem but often the solutions fall into one of two categories. The gamut of one device must be brought into line with the gamut of the other. This can be done by shrinking the input gamut or by clipping it, often termed photographic and solid colour matching, respectively.

Solid Colour matching is essentially clipping the input device gamut. It is useful for spot colours, i.e. colours to be printed with a specific ink, and graphic images. Where ever there is a direct one for one colour match between the devices, that is the colour that is used. This therefore preserves the actual colours in the majority of cases. Where there is no directly equivalent colour, the CMS searches for the best possible match. The definition of best match could vary from system to system.

Photographic matching does not preserve exact colour matches, but does preserve the relationships of colours to each other and to the background reference white. This is based upon the human perception of colour. In general, the brain is very good at compensating for lighting differences. White squares seen in bright or dim lights still appear as white. The brain takes into account the amount of light and automatically adjusts the perception of other colours in proportion.

The CIE colour definitions define the three proportions of each of the colour values in relation to a reference white. The common reference colour space of a CMS does the same. For out of gamut colours, the relationship of the colour to reference white is used, but each value in the triple is adjusted by the same proportion so that the new colour still has the same relation to reference white, but is now within the gamut of the output device. The same proportional adjustment is made to all the colours in the image. Since the eye takes note of the relationship of the colours to each other, it accepts the printed image as similar to the original. However, none of the colours now correspond exactly. This technique is appropriate for photographs, where acceptability is subjective. It is not appropriate where ever an exact match is required.

Some application packages allow the user to select whether to use Solid Colour or Photographic matching algorithms. Some decide which to use on the basis of the image files being imported or the creation of the colours in a colour editor. The latter often provide for the user to override the system choice if preferred.

The crucial part of the Colour Management System is the range of device profiles. Some CMS provide a wide range of device profiles, some provide a generic set, but allow for the plugging in of third party profiles. Some allow the user to create profiles for their own devices. At the moment, what is available with a given CMS may depend on the execution platform.

Generic device profiles have to make assumptions about the calibration of devices. ANSI have produced the IT8 standard, and CMS utilise this. They provide an IT8 image, or images, which is an image of known CIE tristimulus values. Images may be either additive or subtractive. These reference images can be scanned in and the colours seen by the scanner converted into the reference colour space. The difference between these values and the known values for the image, gives the basis of a profile for the scanner. The ease of use of this system means there is no need for a great number of scanner profiles. One or two generic scanners are provided to give a basis for the initial conversion of scanner colours to reference colour space. The user can then use these as a basis for building custom scanner profiles.

Profiling monitors is a much more complex process and systems provide a wider range of profiles for the user to select from. For really accurate calibration a spectrophotometer is required.

Calibrating printers is also a complex task, and the default profiles provided with CMS contain a multitude of printer definitions. All systems provide a generic CMYK profile for use where there is not a specific one provided. It is possible to give the user the option of creating a printer profile, but this is a complex task and the facility is not currently widely available. It involves entering the printer's CIE XYZ settings for a batch of reference colours. These can be obtained by measurement with spectrophotomers or colorimeters.

The final element in a fully-fledged CMS is the provision of open-systems features. This is of relevance when transferring image files between systems. The displaying or reproducing system needs information about the conditions under which the image was originally generated. This is provided by the encapsulation of image origination information within the image file itself. This is achieved by embedding device profiles along with the image. For this to work in an open systems setting a standard for device profiles is required. This is covered a little more in the section 4. Files containing such source information are sometimes known as tagged files.

Back Next Contents