There is no such thing as "accurate" colour in a photograph. The colours of the original are not recorded in any objective fashion, wavelength for wavelength, saturation for saturation. Instead, they are re-created: they are an illusion. Also, there is often a big difference between the colours that were there, and the colours that we remember. 

Misty morning, Chengde. Are these the exact colours of that Chinese dawn? Dunno. But they're pretty much as I remember them; which is to say, as I felt them. Mood is often more important then spectroscopy when it comes to colours that look right. Leica MP, Kodak ISO 100 slide film.

Arguably, the gap between memory and objectivity leads to three levels of colour quality. The first is believable (or barely acceptable) colour. The second is convincing colour. The third is pleasing colour. 

People tend to forget that we see with our brains as well as our eyes, and that to a very large extent, we see what we expect to see. They also forget that (for example) a "blue" sky can be anything from a washed-out grey with just a tinge of blue through to almost ultramarine. If we don't know what colour the sky was in the original, almost anything can be believable or even convincing. The Wikipedia "shades of blue" page provides an interesting array of blues, and of course they can look very different on different monitors. 

Beach, Paarnu, EstoniaThe Baltic can be surprisingly warm. The colours in a picture like this, essentially a happy-snap or holiday brochure image, have little need to be anything more than believable. The blue of that sky would be rendered differently by different sensors and even lenses, let alone different films. Leica M9.

The boundaries of believable colour are quite often explored by bad colour reproduction in newspapers, and by the American NTSC television system, the initials of which are sometimes hijacked as standing for Never The Same Color Twice. Believable colour can also be seen in all but the worst of low-end hand colouring. Convincing colour is what most people and processes achieve most of the time, and pleasing colour is, well, pleasing. Almost paradoxically, the best hand colouring can be both convincing and pleasing. 

River Thouet. Expectations again. We are all used to the greenish light filtered through trees on a riverbank, but if there were a person in this picture, their skin tones might look distinctly unwell. There are priorities in colour renditions, and usually skin tones trump all. Leica M9. 

"Wild" spa. The picture on the left is much more the way things "really" were, with the blue of sky-light and the green influence of the trees, as in the previous picture. But most people would probably find the picture on the right to be more acceptable. Leica M digital.

In a print, the re-creation is done with dyes or pigments. To make life more interesting, it's not even "one dye, one colour". You can create an astonishingly convincing range of colours with just three dyes, each with its own spectral response, plus black. The three-dye or three-pigment process is called CMYK, for Cyan, Magenta, Yellow, blacK (B is already taken for Blue). Because different dyes have different spectral responses, different manufacturers' versions may be more or less convincing, and for the same reason, high-end ink-jet printers and some photomechanical printers "split" the colours and even the blacks: my Epson, for example, uses CMY plus "vivid" C and M, plus three intensities of K: standard, light and extra light. 

Urinals, Cologne. It's a fairly safe bet that the tiles and urinals are white, not green and magenta. On the other hand, fluorescent lighting does not provide a continuous spectrum, and a combination of an ultra wide angle lens (15/4.5 Voigtländer) and a digital sensor can show us something rather different from what the human eye sees. Set the pink tiles to white in Adobe Photoshop and the whole picture goes horribly green.

On top of this, our perception of colour in a print is greatly influenced by the light under which we see it. The human eye is extremely adaptable, and can see "white" as "white" under several shades of daylight (sunlight, overcast, shadowed light) tungsten light and fluorescent light. As we all know, films are not so adaptable, and for digital photography we must either choose one of several settings or rely on automation. This is a further reminder that we see with our brains as much as with our eyes; or possibly more with our brains than with our eyes. 

Stable door. I genuinely don't remember what the lighting was here. Artificial, obviously, and probably "photo quality" fluorescent. It doesn't matter.The blue chair is quite a bit less saturated; the door is a more neutral grey; the stonework less yellow. The overall effect in the picture, though, is a sort of "vivid nostalgia", which is the effect I wanted.

An allied consideration, this time more objective than subjective, is metameric failure: the way that colours look different under different lighting. The original subject reflects one set of wavelengths, but the dyes or pigments in the print reflect another set. The closer the match between the two sets of wavelengths, the closer the colours will remain under different lighting. But if the two are different, then the print may look one colour under (for example) daylight and another tungsten. 

This is often most evident in black and white ink-jet prints from low-end printers: they may well look fine under daylight, but greenish or magenta under tungsten. Re-balance them to be neutral under tungsten, and they will then look magenta or greenish under daylight. An old trick to minimize the problem is to make a print that is slightly warm (brownish) under any light. That way, it's less brown under one and more brown under the other, but never greenish or magenta. The effect is almost paradoxically less obvious in a colour print, but by then we're back to believable, convincing and pleasing, and seeing what we expect.

Moving on to pictures on a screen, the reconstruction is again achieved by three colors: RGB or Red, Green, Blue. The blacK is provided by the unlit background. Again, each pixel has its own spectral response, and different systems likewise offer different spectra: traditional CRTs, LCD, plasma, LED.... One of the big differences between cheap monitors and expensive ones is that as a very general rule, if you pay more, you get better colours. Because the screen is normally self-luminous, you don't have the same question of ambient light colour, but ambient light intensity will have some effect.

Frances filing her nails. Automatic white balance (AWB) can be a great enemy of quality. The red sweater fooled the camera (a Nikon D70, I think) into creating a very blue rendition. I corrected for skin tone using Adobe Photoshop Elements, which is astonishingly effective, but the trouble is that next to the blue version, the corrected version looks overdone. Cover the original with your hand and the corrected version looks fine: another psychological trick.

Now let's move on to the specifically photographic question of exposure. Over quite a wide range of exposure, colours remain convincing without remaining the same. Less exposure gives deeper, richer, more saturated colours; more exposure gives lighter, airier, less saturated colours. Too little exposure and they start to go muddy; too much and they "blow" to a featureless white or sometimes (especially on film) to a rather nasty yellow. Exposure can be especially important in portraits, but it can also matter in just about any type of photography: deciding whether to add a little exposure, or cut exposure a little, is an aesthetic choice. 

Mision de la Purisima Concepcion, Lompoc, CA. This version, as shot on ISO 1000 slide film in a Leica, emphasizes the heaviness and gloom of the religious symbolism and indeed of the furniture.

Increase the exposure slightly, and decrease the contrast slightly, and the mood is very different. This is one reason for bracketing: because you are not totally sure which effect will work best, or maybe even which effect you want.

Saturation is another very photographic question. Throughout the 1960s, 70s and 80s, the saturation of colour prints tended to increase steadily, to the point where many old-fashioned photographers declared that they were "too saturated". This is surprisingly easy to counter. Hold a print up next to the original subject, and almost invariably, the original subject is more saturated. Of course it is entirely possible to produce an over-saturated print, but once again, we are in the realms of aesthetic choice, not accuracy. 

Church candles. With a top-quality modern digital camera, you could make this look as if it was shot under white light and (with the help of Adobe Photoshop or something similar) give her a convincing skin tone. The result would be awful. We need the murk for the mystery. Leica M digital.

Equally photographic is the question of colour balance in the final image. As the late, great Terence Donovan once phrased it, "Have you ever had a client complain because a picture is too ****ing warm?" In portraits in particular, "warmth" (a bias towards yellow and red, though preferably not magenta) is likely to be a lot more acceptable than a bias towards blue, green or blue-green (cyan). If you could directly compare the subject and the picture, under identical lighting, you could probably adjust the colour in the print until it was quite close; but again, because it's a reconstruction, not a record, you would probably not be able to get all the colours spot-on. 

Escalator, photokina. What colours are you expecting here? The light is a mixture of fluorescent and tungsten, so any rendition that was truly objective would look awful. As it is, I find this quite pleasing and convincing. Leica M9.

Exposure, saturation and colour balance all bring us back to our three levels of colour: believable, convincing and pleasing. To be sure, you can use other words: adequate, acceptable, good, pleasing, nostalgic and more; but none of them refers, for a moment, to accuracy.

Finally, for pedants and for devotees of the unusual, it is worth mentioning that the Lippmann interference process is sometimes held out as objective; but anyone who has ever seen a Lippmann photograph will have their doubts as to how meaningful this is. Equally obscurely,  Capstaff's two-colour Kodachrome process (1915) produced astonishingly convincing colours from just red/orange and blue/green. Edwin Land also worked on a two-colour process from the 1950s onwards but unfortunately I have only ever found fragmentary references to it: I have never looked up the original 1959 Scientific American article. He based his Retinex theory of colour vision on the phenomenon. 

Go to Photography

Go to Index

Go to Home Page

Words and pictures copyright (c) Roger Hicks 2016