You are here: Home > Still images > Colour Management in Practice
Colour Management in Practice
Last updated: 15 December 2008
Published in:
Digitising analogue media |
Creating new digital media
Tags:
colour management |
workflow
Summary
This document provides a detailed explanation of the different approaches to colour management. The document also covers the process of calibration of monitors, scanners and printers as well as conversion between aditive and subtractive colour systems.
Contents
- Introduction
- Why Colour Manage?
- Traditional Colour Management
- Modern ICC Colour Management
- ICC Profiles
- Rendering Intents and Gamut Mapping
- The International Color Consortium
- Conclusion
Introduction
This paper provides an overview of current colour management practice within digital imaging today, especially as it pertains to the workflow of a large digitisation project. As such, it largely reviews the workings of the ICC colour management systems that are now established as the standard methodology for producing accurate colour within today's colour workflow.
This current paper follows on from another JISC Digital Media advice document: Colour Theory: Understanding and Modelling Colour. JISC Digital Media's Colour Theory paper provides an introduction to the science and history of colour and its representation. Colour management is not an easy subject and a little understanding of the supporting theory can go a long way to help you come to grips with the intricacies of colour management in today's workflow. Another advice document of relevance is Colour and Resolution Targets which deals with the characterisation and calibration of imaging hardware such as scanners or monitors.
Why Colour Manage?
The need for colour management is based on a simple dilemma that anybody working with colour images has quickly discovered. Put simply, the problem is that it is very hard to make a digital image look the same across a range of different output devices; the same image may look quite different when printed on different printers or for that matter when viewed on different monitors. When trying to provide consistency across a range of different output devices such as printers, monitors and projectors the task becomes very hard indeed.
Colour management is required simply because every different device produces or reacts to light differently. Each device therefore needs to have different colour values to hopefully produce similar results. The goal of colour management therefore is simply to try and provide a system that guarantees that images look the same where ever and when ever they are viewed.
It is easy to think of colour management as being something new that has popped into our reality over only the last few years, but in fact magazines and books have been full of high quality and well-managed colour for many years before we started creating and managing images ourselves. This is because colour management has traditionally been the realm of the print industry who have needed to produce high quality colour in all the printed matter for public consumption.
However, for the majority of digitisation projects, the main delivery goals are based around using the images within a 'learning and teaching' scenario and the majority of image use is on the monitor although some will indeed be in print. As there is no one method of capture (camera, scanner etc.) nor one method of output (Web, print, MS PowerPoint) the digitisation project has to consider the requirements of colour management in a much more versatile context than was traditionally necessary.
Traditional Colour Management
The first efforts to control the quality of colour reproduction were made by the print industry who based their workflow on the analogue systems that had been in use for the previous 40-50 years.
In this workflow all the equipment was professional, well established and very expensive. All devices were calibrated and the colour controlled by evaluation of test images by a colour expert. In this process, a test transparency, of known and standard quality was scanned and then the image was taken through the whole process until printed. This test print was then compared to the original standard image by skilled operators. Who, having viewed the image, adjust the controls of the various devices (scanner, printer etc.) until the colour of the output print matches the colour of the original transparency. Depending on the skills of the print operator this iterative process of 'print-view-adjust-print' is able to 'close-the-loop' and provide colour quality of the highest order.
Of course it is still necessary in this workflow to make sure that all equipment is completely and accurately calibrated whenever it is used. This has to be done at regular intervals to make sure that each device is working within its set limits and has not been allowed to drift away from standard settings.
In this traditional print-based system there will of course be a transformation at some stage from the RGB colour data that is captured into the CMYK data that is printed. This could be undertaken at various stages in the workflow, but traditionally it was done at the earliest stage possible, even within the scanning process. This is why some older scanners are known as 'CMYK scanners' for the only output that is available is CMYK. This does not mean that the scanner captured directly into CMYK but rather that it captured in RGB and then immediately and automatically transformed the image into CMYK.
Device dependent workflow
This traditional system of colour management is completely dependent upon the specific devices that are used within that workflow. If a device is changed or loses its calibration then the accuracy of the whole image workflow becomes suspect. It will work and work well, but only within an established and known workflow and only if the whole workflow is used in its entirety. Images must be introduced at the beginning of the workflow, normally as a transparency that is scanned to produce the image that goes through the workflow to print. It is not however possible to use a digital image that has been created from an external capture source within this flow, or indeed to take an image that originated within this flow and use it somewhere else. For this reason this workflow is often called 'closed loop' colour management.
In early days of digital imaging this often lead to the print industry appearing to be inept or unhelpful to those working with digital images. However, the only safe way of bringing an RGB digital image into the workflow was to first burn it back to a transparency and then scan that!
It is quite possible to create a colour transform that will standardise this simple workflow:
However, the problem arises when we want to change this workflow in any way. The transformation is dependent upon the devices used within the workflow and therefore if we add a new device we will need to create a new transformation for that particular workflow and each new device will demand the creation of a new transformation for that particular set of devices:
This can quickly become very confusing and unsustainable within a workflow with a growing number of new devices being introduced. On the other hand, it is undeniable that this 'device dependent' workflow can provide a high quality colour management system within an established colour workflow such as the traditional world of print.
As the world of digital imaging grows, so there is an increasing demand for a system that allows for a variety of input and output devices to be used within a colour managed workflow. We therefore need to create a system that is 'open loop' and not directly based upon each device within the workflow.
Modern ICC Colour Management
Device independent workflow
This workflow greatly simplifies the system by using a standard intermediary colour space from and to which all other devices can be transformed:
With this colour management system any device can be used as long as it has been calibrated and characterised to create a profile that compares it to a 'standard' colour space. In ICC colour management this is called the 'Profile Connection Space' or PCS and is based upon the CIELAB (CIE L*a*b*) or XYZ colour spaces (see Colour Theory for further explanation).
So at its most basic, the ICC workflow will require a minimum of two colour profiles within it to work properly:
Although there are two profiles associated with this transformation, they are both used by the Colour Matching Module to create one transformation which simplifies the process and helps avoid errors.
Viewing the image within the workflow
This system will provide high quality colour for the transformation of the image from capture to output, however we also need to provide a method for allowing the image to be viewed and reliably edited within the workflow. This is done by creating another profile to control the transformation of colour from the PCS to your monitor.
Whilst the image is held within the workflow, it is kept within a 'working colour space' that is a virtual space (it does not represent any device). This 'working colour space' could match the PCS and use the CIELAB colour space, but this is a very large and unbounded colour space that is far bigger than is required for output to either a RGB based monitor or any CMYK based printer. It is therefore more normal and better practice to use a colour space that provides a closer fit for the intended use of the image.
When an image is viewed on the monitor, two profiles are again required. The colours in the workspace need to be transformed into the monitor colour space, so both a profile for the monitor and a profile for the colour workspace need to be used to make the transformation. The profile for the monitor will be made by calibration and characterisation, whilst the profile for the 'working colour space' is a known quantity defined as part of the 'working colour space'.
The actual transformations are undertaken by the Colour Management Module or CMM, although also known by a range of names including the Colour Matching Module, Colour Management Engine or Colour Matching Method. Both MS Windows and Apple ship with CMMs (ICM for Windows and ColorSync for Mac), however only Apple offers this at operating system level. For Windows it is normally recommended to use the CMM that comes with your imaging editing program such as Adobe Photoshop
Colour workspaces
The choice of 'working colour space' will depend on how you intend to use the image in the future. It might be considered a good idea to save the image in a form that in no way compromises the future colour quality of the image by choosing a very large colour space that will easily contain all the colours within the original image (although this normally needs a larger colour bit depth to make the most of it). But, in normal practice (especially if working in 24bit colour) it is better to choose a colour space that provides a closer fit to the colours that are actually realisable. In other words, there is no point in holding colour information if it can never be output.
The choice of 'working colour space' will also be important if any images are opened which have not up to that stage been colour managed. By converting them to the 'working colour space' it will be possible to view them knowing that they can be safely edited and output with reliable colour.
There is a large range of available colour spaces, of which the most common are:
- sRGB IEC-61966-2.1
This is a standard colour space originally promoted by Microsoft, Hewlett Packard and some other PC companies. It is based on the expected quality of an 'average' consumer (2.2 gamma and D65 white point) PC monitor. It is the standard colour space expected by all Internet Web browsers. It has now also become the standard colour space used by all consumer based digital cameras and scanners. So although it was designed to be a 'lowest common denominator' colour space with only a limited colour gamut, it is still the best choice for all images that are to be viewed on the Internet, or in any situation where the quality of the viewing monitors is low or unknown. On the other hand it would be a very bad choice for any images that are to end up going to high quality print. - Adobe RGB (1998)
Adobe have established this colour space (originally based on SMPTE-240M - a standard for HDTV) as a recommended standard for images that are to be converted at some stage to CMYK for print. It is likely to shortly become an ISO standard and has been accepted by most users as providing the best overall compromise between quality and gamut size. - ProPhoto RGB and Wide Gamut RGB
Both these colour spaces have very wide gamuts that include a large amount of colours that can be neither printed nor viewed on most monitors. However they do contain most of the extended gamut available to photographic materials and high quality desktop printers. So although this amount of information might be considered overkill for most uses, it can really improve a high-quality image that is to be printed onto photographic materials. As these colour spaces are so wide, it is normally advisable to only use colour images that have at 16bits per channel, to make full use of the colour space. - ColorMatch RGB
This is an open standard working space based upon the gamut of the high quality Radius Press View monitors. It has a gamma of 1.8 and is a well established working space for Mac users doing pre-press work. It is a known standard and if working with an established print-orientated workflow, it might be necessary to work with ColorMatch RGB, but it is unlikely that anyone starting afresh would choose to use it.
ICC Profiles
ICC profiles are the building block on which all ICC colour management is based, however by themselves they are nothing more than very simple 'look-up tables' that record the differences between a device's intended colour and its actual created colour. In other words they allow the transformation of colour values to/from the device and the Profile Connection Space (PCS).
| R | G | B |  |
L | a | b |
| 255 | 255 | 255 | 100 | 0 | 0 | |
| 255 | 255 | 240 | 99 | -2 | 7 | |
| ... | ... | ... | ... | ... | ... | |
| 0 | 0 | 20 | 1 | 2 | -9 | |
| 0 | 0 | 0 | .58 | 0 | 0 |
ICC Profiles can be thought of as paired lookup tables with columns of numbers for each colour in the colour space.
Any colours not explicitly listed in the table above are obtained from interpolation by the Colour Management Module.
There are a number of profiles that may be required, each made for the transformation of colour to or from the PCS.
| Type | Use |
|---|---|
| Input Profile | Scanner, Camera etc |
| Output Profile | Printer, Film Recorder |
| Display Profile | Monitor (CRT, LCD), Projectors |
| Device to Device Profile | Custom Transform from Device to Device |
| Colour Space Profile | Transform to or from a Working Colour Space |
We will look at the creation of just the first three of these types of profile:
Input profiles
An input or 'capture' profile is made for all devices that are making an original digital image from some other analogue source, such as a digital camera or scanner.
The device is characterised by capturing a 'test target' with a range of established and standardised colour values. Each one of the colour patches on the target has a known value that is contained in an associated data file that accompanies the test target. First the test target is captured and then each one of the colour patches within the image of the target is compared against the corresponding value for that colour within the data file.
The IT8.7/2 colour test target
From this a long list of differences is created which can be used to create a custom adjustment for each of the colours within the target. The more colours the target has the more accurate this system will be. Colours that fall between those on the target will have to use an averaged reading based on the closest colours. The standard test target for creating capture profiles is the IT8.7/2. In its standard form this has 240 test colour patches although some manufacturers make versions that have a few extra patches.
Of course this whole process relies heavily on the data files being an accurate representation of the colours on the target. The data files can be made in a number of different ways that will control both the accuracy of the data and also the cost of manufacture:
- Generic 'Target-Data' set: In this case the target is printed to an established value and the accuracy of the match will depend entirely on the accuracy of the calibration of the printer as it endeavours to print to the established values within the data-files. How accurate this is will depend entirely on the quality of the printing. In general these are not very expensive but they also tend to not be very accurate either and are typical of the 'manufacturer-supplied' target-data set.
- Batch 'Target-Data' set: To provide a more accurate set of data-values for the colour patches it is possible to take one of the printed targets from each production batch and individually measure the colour values on the target. As there is only a slight variation of colour within each batch, this should provide a more accurate set of data-values for that target and therefore a more accurate profile. These are the most common target-data sets supplied for the creation of quality profiles.
- Custom 'Target-Data' set: The most accurate profile will be made by individually measuring the colours of each and every colour patch and creating a custom set of data values specifically for that colour target. These target-data sets will provide the highest quality but at the highest cost.
It should be realised that the more accurate your target-data set is, the quicker it will become inaccurate if the target is allowed to fade or become marked. It is therefore imperative with all targets that they are looked after and kept away from light or unnecessary handling. However careful you are with your targets, they will in time fade and it is then best practice to either replace them or to at least re-measure them and create another data-set that matches the 'faded' target.
Output profiles
Output or 'printer' profiles are made in a similar way to input profiles but in this case the test image is in the form of image-data of a colour target in which every colour patch is of a known and standardised colour. This test image is then printed using the device, ink and media that you wish to profile. The print of the target image is then 'read' by a colour photo-spectrometer that measures the colour value of each of the printed target patches. When this data has been recorded into a file, the 'measured' values can be compared with the 'correct' values and again a look up table is created which provides a correction for each colour in the test image.
With printer profiles it is important to realise that 'any' change to how the print is made will make the profile less or even totally ineffective. The profile will be as dependent upon the type of paper and of course the ink as it is on the printer itself. It is therefore important to standardise on the ink and paper used by your printer and if you need to use more than one type of paper, then of course it will also be necessary to make and use a different profile for each paper.
As printer profiles have to be made for each printer-paper-ink combination, it is very hard to make any generic profiles for printers and this is why manufacturers' profiles tend to not be very accurate. Also as each profile has to be custom made, they can tend to be relatively expensive. However, expense is comparative and for any printer that is relied on to produce high quality colour or to provide a relatively high throughput, the costs might well be considered very low.
Monitor profiles
It is easy to understand that an input profile is comparing the colours created by that device with the Profile Connection Space (PCS) and that an output profile is again comparing the colour created by that device with the PCS, but what exactly does a monitor profile do?
A monitor profile is made to compare or 'characterise' the colours created and shown by the monitor against those in the PCS. Calibrating the monitor is a relatively easy task and yet it is perhaps the key part of any colour management system. Unless you are quite sure that you are looking at an honest and high quality representation of your digital images, it is quite impossible to make any colour quality judgement or undertake any editing.
Until recently most users calibrated their monitors using 'subjectively' based software programs such as the Adobe Gamma program that came with Photoshop or slightly more advanced programs of the same nature such as Sonnetech Colorific (now ColorWizzard). These programs created a monitor profile by asking the user to make a series of subjective choices where they would alter the gamma and colour settings to match a known standard output on the screen:
This system might well be considered better than nothing but it is hardly an accurate or objective method of characterising a monitor. A quick test will easily show that different users will calibrate the monitor to different values, highlighting the vagaries of this kind of subjective and user based system.
It is therefore not surprising that the importance of this part of the colour workflow has lead to much re-consideration on what is the best way of undertaking this critical characterisation.
To fulfil this demand we are now seeing a range of more objective and automated systems becoming available.
These consist of a colorimeter or colour photo-spectrometer that is attached to the monitor, which then takes readings of a range of colours that are provided by the associated software. There is a large range of these monitor-calibration systems at an equally large range of costs.
Like many other areas of technology, you get what you pay for and, even though the cheapest are likely to provide better results than the Adobe Gamma system, it will be worth paying for the best if you are trying to provide the highest quality and most reliable colour.
It is important to remember that the phosphors of a monitor will change over time and therefore it is important to regularly calibrate your monitor, certainly at least once a month is recommended.
Rendering Intents and Gamut Mapping
The colour 'Gamut' for a device is the name given for the full range of colours that can be recorded or output by that device within it's colour space. The exact gamut of any device will also depend upon the media and how it is being viewed or lit. Normally, in general, a CMYK printer device colours pace will be smaller (less saturated and fewer colours) than a RGB capture device colour space.
A typical RGB colour space
A typical CMYK colour space
CIE Colour Gamut
The full colour gamut of the human eye is described by the colour area. The blue and the yellow line describe the estimated colour gamut of a RGB and a CMYK device.
When a transformation is made from one colour space (normally the capture colour space) to the other colour space (normally the output colour space); the colours will need to be mapped from one gamut to the other. If the gamuts have significantly shrunk (or indeed grown, although this is less likely) there will be some colours that are attainable on one device whilst not being realisable on the other. So the Colour Management Module (CMM) will need to change some colours as they are mapped to make the most out of the gamut of the new colour space.
The ICC has defined four different ways of rendering this gamut mapping depending upon the intention of the user. These are four Rendering Intents, called Perceptual, Relative Colorimetric, Absolute Colorimetric and Saturation:
Perceptual rendering intent
In perceptual rendering intent, the whole gamut of the original image is compressed until it fits within the gamut of the destination colour space. This has the distinct advantage in that it keeps all the relationships and differences between the colours within the image. This means that no colours are clipped, leading to less likelihood of any banding in blocks of colour as they fall out of gamut (as can happen with fully saturated blue skies). Although this rendering intent is normally the first choice for all photographic images it does however have the problem that all colours within the image may well be changed in the process.....so although it can 'look' good, it is not likely to provide the most 'accurate' colour.
The perceptual rendering intent is normally considered the best choice for full tone photographic images where it is the intention to retain the overall 'feel' and 'look' of the image. It is certainly best if there is a large amount of 'out of gamut' colours that need to be changed for the image to fit into the destination output. It is favoured for any 'automatic' system or as a 'standard' across a number of images, where it can be relied upon to provide a reasonable quality.
A typical RGB colour space
A typical CMYK colour space
Perceptual Rendering Intent
Aims to preserve the visual relationship between colours in a way that is perceived as natural to the human eye, although the colour values themselves may change.
Suitable for Photographs
Relative colorimetric rendering intent
Relative colorimetric leaves all colours the same when they fall within the gamut but those colours that fall outside the gamut are simply mapped to the nearest 'in gamut' equivalent. So although this may well leave the bulk of colours within the original image untouched and therefore hopefully accurate, it will unfortunately clip all out of gamut colours. The typical place where this can be unfortunately visible is within bright blue skies; using a colorimetric intent will leave all the 'out of gamut' blues exactly the same colour as a 'just in gamut' blue. This can create visible and very ugly banding in the sky.
With relative colorimetric rendering intent the black and white point are mapped to the new colour space, making sure that the 'greys' appear to remain grey even if this leads to some slight movement of colours within the rest of the image.
Although relative colorimetric has the danger of clipping some highly saturated colours during the transformation, if there are very few of these or they are in inoffensive areas of the image, it can provide a better transformation with a much higher percentage of colours remaining unchanged thereby providing more accurate and brighter colours. However, to get the best out of the relative colorimetric rendering intent does presume a high level of knowledge and skill on the part of the operator. So whereas it can be seen that 'relative colorimetric' rendering intent 'can' provide higher quality than 'perceptual' it also presumes a higher level of operator skill to get it.
A typical RGB colour space
A typical CMYK colour space
Relative Colorimetric Rendering Intent
Like Absolute Colorimetric, out of gamut colours are moved to the closest in-gamut colour that will preserve lightness and hue but not saturation. However in this case the white point is adjusted during the transformation.
Suitable for photographs where only a few colours are outside output gamut, an experts choice
Absolute colorimetric rendering intent
The absolute colorimetric rendering intent works in much the same way as the 'relative colorimetric' rendering intent but no adjustments are made to the black and white points in the transformation. This means that all colours within the original colour gamut are mapped exactly to the same colours within the destination output. This rendering intent is useful for transforming any images that contain important colours that must be exactly the same in the new image, such as 'signature' or 'trade' colours such as the Fuji 'green', the Kodak 'yellow' or the McDonalds 'red'.
A typical RGB colour space
A typical CMYK colour space
Absolute Colorimetric Rendering Intent
Identical colours are left alone. Out of gamut colours are moved to the closest in-gamut colour, meaning some colours will become identical. The white point is not adjusted.
Suitable for images with signature colours, or anything where accuracy is more important than visual appearance
Saturation rendering intent
The saturation rendering intent preserves the relative saturation of the 'out of gamut' colours as they are mapped from one colour gamut to the other even at the expense of having accurate hue and lightness. This rendering intent is normally used for business graphics where bright and fully saturated colours are more important than absolute colour accuracy.
A typical RGB colour space
A typical CMYK colour space
Saturation Rendering Intent
All colours are transformed to the brightest saturation possible. This provides vivid colours but does not necessarily provide colour accuracy.
Suitable for graphics
The International Color Consortium
The International Color Consortium is an industry based group formed in 1993 with eight original stakeholders that included Adobe, Apple, Microsoft and Eastman Kodak. Its intention is to encourage the creation of a standardised, open, non-proprietary and cross-platform system for high quality and accurate colour management.
As we have seen the ICC approach to colour management is largely based around the use of colour profiles to objectively quantify the colour performance of the device for which it is made. This work was based on the previous use of colour profiles by ColorSync (Apple), Kodak, Adobe and others. The ICC profile was adapted from the ColorSync profile that had been developed by Apple. By 2004 the ICC profile was into its 4th version with widespread support from all industry players working with high quality colour.
Since 1993 support has grown for the ICC's approach to colour management and the ICC now has cross-industry support with over 70 members. The ICC system has been accepted as the basis for ISO standardisation and is already integrated into all Adobe products with most other software quickly falling into line. On the other hand this number of members has sometimes led to problems with agreeing on the wording for a standardised specification and the clarity has suffered. The ICC is working on this and will continue to improve and update the ICC specification.
Further details of the current state of the ICC can be found on their website.
Conclusion
Colour management is a topic that has a reputation as being difficult to really fully understand and even harder to reliably implement. However we are now beginning to move towards complete standardisation in colour management systems and ICC colour management is fully supported by Adobe, Microsoft and Apple with the rest of the digital imaging community following fast behind. This should quickly lead to a more automated and seamless approach to colour management.
The most important requirement to getting more reliable colours for all of us, whenever we use or view an image, is that we all take some effort to understand and implement ICC colour management within our capture and delivery workflows. At the very least, all users from project staff through to Web browser need to be aware of the requirement for ICC colour management and be willing to take the necessary steps to apply colour management to the images within their workflow even if this is just to make sure that all monitors within the workflow are regularly calibrated.
ICC colour management is no longer hard to implement, but it does require that all sections of the image workflow are included within the system. The ICC system will only work if everyone is using it and of course can do little for the workflow or project that has decided not to use it.
Last updated: 15 December 2008
Published in:
Digitising analogue media |
Creating new digital media
Tags:
colour management |
workflow
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
