Profiling

Practical colour profiling

By Bob Newman, first published June 2013

There is nothing at all wrong with a workflow which starts and ends with the manufacturer’s processing. However, some people want or need to take more control of aspects of their workflow, particularly colour management. There may be many reasons for this, for instance, a requirement for consistent results across a range of equipment, possibly from different vendors. Others have developed their own colour style and want transferability to new equipment. Still others are working for clients who demand compliance with in-house standards. For whatever reason one chooses to take the plunge into colour management it is worthwhile taking it systematically, it is one area where trial and error can take a very long time to produce good results.

The basics of colour

The basis of colour in photographic reproduction is the application of notionally separate stimuli to the three different colour receptors in the eye, the ‘cones’. Each of these has a different response to light across the visible wavelength range, as shown in figure 1.

Figure 1: The response of the CIE XYZ primary colours, compared with the response of the three types of cone in the eye, L (red), M (green) and H (blue). The XYZ space also accounts for the perceptual effects of the rods, the luminance detectors in the eye.

Essentially, each colour receptor gives a single response which says, ‘the total amount of light in the frequency range is this much’. Thus, we judge colour as a combination of three values, which are conventionally called ‘red’, ‘green’ and ‘blue’. Any distribution of light spectral power that produces the same values for ‘red’, ‘green’ and ‘blue’ will look to be the same colour, no matter if they are very different in actual makeup. Therefore, the job of colour photography is to ‘record’ those three values (known as ‘tri-stimulus’ values’) using a camera, and then create an equivalent stimulus for the viewer, with whatever output device is used.

Colour spaces

A set of colours each defined by three separate values defines a ‘three-dimensional’ space. Imagining that the three stimuli form the edges of a cuboid, we can see that the size of our ‘colour space’ is defined by the ‘size’ of the edges of the cuboid. By assigning a set of numbers to represent a range of stimulus values, we arrive at a ‘colour model’. The total range of different colours recordable is called the ‘gamut’.

Another interesting mathematical property of this three-dimensional representation is that the direction of the three axes does not affect the ability to define separate colours. Thus, mathematical transforms can be performed to rotate the axes in different directions, so the three values we use do not have to represent ‘red’, ‘green’ and ‘blue’. In fact, there are advantages in doing so. One of the properties needed for image editing is that changes in one colour ‘channel’ cause a predictable effect, for instance, making the image ‘redder’ always has a reddening effect, which will not be the case for all colour spaces. The formative work on colour undertaken in 1932 by the International Commission on Illumination (CIE) defined standard colours in terms of an ‘XYZ’ space, where ‘Y’ corresponds to the ‘luminance’ function, thus being usable in one-colour (monochrome) systems to give an accurate representation of apparent brightness. ‘Y’ is in fact close to the eye’s green response. A more predictable variant of XYZ is ‘CIE Lab’, where ‘L’ is ‘lightness’ - broadly corresponding to luminance or ‘Y’ and ‘a’ and ‘b’ are ‘opponent’ channels, green-magenta and blue-yellow. Lab has a huge gamut, larger than the colour capability of the eye, and is thus a useful colour space in which to perform colour manipulations without losing information. Other commonly used standard colour spaces are sRGB (the space used in the web and for most electronic reproduction), Adobe RGB (designed for print reproduction) and ProPhoto RGB. Of these three, sRGB has the smallest gamut and ProPhoto the largest (though not as large as Lab).

Colour workflow

We can now begin to understand the role of colour spaces in a real-life work flow. End to end, a photographer will work with at least three colour spaces. Two of these are arbitrary, defined by the physical operation of the capture and output devices. Here the ‘direction’ of the axes in the space are determined by the characteristics of dyes, used for the colour filters in the camera or LCD display, or the inks in the printer. At this stage, it is worthwhile noting that if you change a printer’s inks, you change its colour space. The third space will be a defined, standard space used in the processing and post-processing. Three is the minimum, but there may be more – for instance one might wish to preview a printed image on a computer screen, in which case both devices’ colour space need to be included. Moreover, a normal workflow will include transformations between colour spaces in processing.

These transformations are made with reference to a standardized colour space, typically CIE Lab. Transformations are made using a mapping from values in the source space to values in the destination space, usually called a ‘colour profile’. So, a transformation from a camera’s native RGB space to, for instance, Adobe RGB will often be made by a transformation to CIE Lab, followed by a further transformation to Adobe RGB. CIE Lab is used as a ‘profile connection space’. The reason for using Lab is that its gamut is so large that it includes most commonly used spaces. If a colour in the source space does not exist in the destination space then the issue must be resolved, and the way it is done is with respect to a ‘rendering intent’, usually ‘relative colorimetric’ or ‘perceptual’. In relative colorimetric, the colours that are out of gamut in the destination space are simply mapped to the nearest in gamut colour, so it’s like clipping the larger gamut to the smaller one. Conversely, perceptual shrinks the whole of the larger gamut to fit the smaller, moving all the colours but maintaining their relationships to each other.

Profiling

The alert reader will have realized now that most of the workflow uses well defined colour spaces, with already defined profiles to move between them. If you are content to go with manufacturer’s preferences, cameras, printers and monitors are the same, they come with the manufacturer’s profiles to translate to and from the standard colour spaces. However, if you want to take control yourself, and ensure a consistent colour response between these devices, or even something different from the manufacturer’s preference, you can do it yourself. In outline, here’s how.

First you need to create an International Camera Consortium (ICC) colour profile for the device. You will need some tools, both hardware and software to do this. To profile camera, the hardware you require is a colour chart of known colours – such as the well known Gretag MacBeth chart (Figure 2).

Figure 2. Colour profiling a camera requires a standard colour target, such as this Gretag Macbeth chart.

Since the colour of the light affects colour rendition you will need to make a profile for each light source you use and take care that the chart is evenly illuminated by this light source – this is indeed one reason for custom profiling, to get an exact profile for your light. The next step is to process that image with a colour managements system (open source ones Argyll CMS (http://www.argyllcms.com/) and LPROF CMS (http://lprof.sourceforge.net/), commercial ones include ColorEyes or for Adobe users DNG Profile Editor) following the instructions for your chosen tool. The result is an ICC profile (or Adobe proprietary one), which you can use with your raw conversion tool of choice, loading it as the default for that camera/light combination. Again, follow the instructions for your tool of choice.

Accurate profiling of monitors requires a tool to read the colours emitted by them. Popular tools include XRite ColorMunki and DataColor Spyder (Figure 3).

Figure 3. Accurately profiling a monitor requires the use of a device to measure the colour output of the monitor.

These are a significant expense, but for soft proofing, an unprofiled monitor is not very useful. The tools come with software and instructions, so simply follow those.

Profiling your printer is best done using a postal profiling service. Ideally, you should produce a new profile for each ink set and paper type that you use. The company providing the service will give a standard test file to print on your printer (using you preferred set of inks and paper) and you post the resultant print to them. They then scan it on a calibrated scanner and produce an ICC profile for your use. With profiled cameras, monitors and printers, you can be sure that your workflow will produce consistent colour results.