There are many differing ways of transforming the three tristimulus values representing a colour, with each transformation resulting in a different colour space. Whatever transform is used, the colour value is represented by three numbers. These tristimulus values are the magnitudes of three primaries that have a defined chromaticity, measured in a particular context, with illumination of a known level and spectral balance. If any of these differ, so will the colour appearance. Whether or not the resultant difference in colour appearance is important to the viewer depends on the application. In some applications, an identical colour appearance is necessary; in others, an approximately correct, but pleasing, appearance is sufficient.
Professional graphic designers working with colour images require an on-screen im- age to accurately reflect the colours that will result when the image is printed. For this to be possible, the characteristics of displayed colours must be precisely defined. For identical colour appearance, the on-screen image must not be desaturated or altered by reflected ambient light. It is therefore necessary to control both the level and the colour temperature of the ambient illumination, as is evident here:
“The Adobe RGB (1998) color image encoding is defined as an encoding of the color appearance of an image that is being displayed on a reference color monitor in a reference viewing environment.
NOTE The intended color appearance can be reproduced exactly on a physical device in an actual viewing environment, only when the actual viewing environment ex- actly matches the reference viewing environment. . . When measured, with the mon- itor turned off, at the monitor faceplate, the ambient illumination level shall be equal to 32 lx. . . . The ambient illumination shall have the same chromaticity as the white point of the display.” – Adobe Systems (2005)
Few computer users require such precision in the reproduction of colour. As long as the colours displayed are sufficiently bright and colourful, and approximately correct, not only will they will be happy with the image, they won’t even realise the colours are not “correct”.
Empirical evidence suggests that strict control of viewing conditions is not too important for normal day-to-day use of on-screen colour. Colour schemes – sets of colour intended to be seen together – are developed by users all over the world and
designer
viewer
RGB values
Figure 2.12: The sources of variability in colour scheme creation include both the de- signer’s and the viewer’s preferences, equipment and viewing conditions. Nevertheless, colour schemes can be created and disseminated, with a large degree of agreement as to their visual appeal, using only triplets of RGB values.
seen by others, in different countries using different equipment and different viewing conditions (fig. 2.12). Yet, by and large, there is widespread agreement on the visual appeal of harmonious colours, as witnessed by the popularity of particular themes and colour schemes on the previously mentioned sites for theming web applications (sec. 2.2).
Typical users, apart from adjusting the brightness and contrast (often to the max- imum setting), adjust neither the colour settings on their display nor within their computer. The colours displayed by their monitors are therefore only approximately correct. This is true even on displays that have been calibrated and for which a colour correction profile exists. The viewing conditions must be tightly controlled for correct colour appearance, and this occurs rarely.
2.5.1 Colour constancy and adaptation
Adaptation in the human visual system is the reason for the non-criticality of the displayed colours. The user still “see” the same colours, even if the reflected spectra of physical objects has changed significantly as the spectrum of the illumination changes. Unless the change in the spectrum of illuminating light is extreme, a human observer is unlikely to notice.
The human visual system has several mechanisms to stabilise the apparent colour even though the wavelengths from a scene can vary widely. Part of this is the ability to “discount the illuminant” – effectively renormalising the visual system so that, after a period of adaptation, the colours appear as expected. Although full adaptation appears to be limited to illumination in the range 5000K-6500K (Poynton, 2003), incandescent lighting (∼2800K) appears acceptably close. It has a slight, but not objectionable, yellowish tint. The spectrum of the light from the setting sun is very different from that at midday, but the colours of many objects do not appear to change. Another cognitive mechanism appears to preserve the expected colour for objects with well defined colours and work “from memory”, rather than what is actually being seen:
“in everyday life we are accustomed to thinking of most colors as not chang- ing at all. This is in large part due to the tendency to remember colors rather than look at them closely.” – Evans (1943), cited in Fairchild (1998).
Hubel (1999) makes the interesting point that although the spectrum of light at sunset almost exactly matches incandescent lighting, the visual impression is quite different. The impression of golden light at sunset does not happen when viewing objects lit with incandescent lighting. In a scene lit with incandescent lighting, all the illumination, both direct and reflected, is derived from the same source, whereas sunrise and sunset scenes have two lighting sources: the golden sunlight, and the sky. The sky has a colour temperature in excess of 10,000K and provides much of the shadow illumination. As the sun sets, the low angle shadows form a larger fraction of the total scene, providing a reference other than objects illuminated by the setting sun. This prevents complete adaptation and gives rise to the golden evening scenes.
Interestingly, this adaptation, if made apparent, can be disturbing. In a house lit with incandescent lamps, the placement of a single 6500K lamp in a bedroom was found to be particularly disturbing to the room’s occupant17. Having acclimatised to the incandescent lighting in the remainder of the house, they walked into their bedroom and turned on the light: the room appeared to be lit with a ghastly shade of bright blue. Yet the same “daylight” colour temperature lighting is widely used in business and workshop environments without any suggestion of a blueish tint, due to the uniform lighting (all 6500K) and the user’s adaptation.
In the same way, colours on a display can vary over a wide range and yet be perceived by users as perfectly acceptable. Graphic designers typically have the white point of their displays set at 5000K. This gives a white that appears yellow when compared with the white on a display set to a 6500K white point, and the white from a 9300K setting (another common option (Zuffi et al., 2007)) will appear to have a bluish tinge. Yet, after using any one of the three for a few minutes, its white will appear pure white, irrespective of any initial impression. This adaptation enables users to feel comfortable using monitors with widely-varying white points, even though, in an absolute sense, the colours are not accurate. This adaptation is also noticeable in the variety of colour settings found to be acceptable on television sets.
The system envisioned in this thesis is intended to produce harmonious colour schemes with a pleasing appearance for users in atypical viewing environment. Typical users do not have calibrated systems nor reference viewing environments. It is therefore essential that any colour schemes created using such a system are both usable and aesthetically pleasing when viewed in a typical non-reference environment.