Chromatic Adaptation and Colour Constancy

Including the approach of Land & McCann

Version date 2024-02-19

VIDEO https://youtu.be/VfKSMU_wcog

 

PROLOGUE: THE BLUE HOUR

I am sure you are familiar with the situation when you are in a brightly lighted room and a neighboring room looks entirely dark. Going closer to the opening, it lights up a little ... and after you have entered, you soon find it sufficiently illuminated.

This effect is called adaptation and is a necessary power of our vision, in order that we shall be able to see clearly under the strongly varying natural situations of illumination -- ranging from a shore in bright sunlight at noon to the shadow under the trees in a forest, on a day with heavy clouding. But there is more to this ...

*

Indeed! - It looks like the recent snowfall has made the garden beautifully white.

Later on, towards the evening, when I have switched on lamps to get comfortable indoor illumination, an accidental look through the window shows me the garden in deep blue light. This is the "blue hour" when there is still a faint remain of daylight outside.

Going outside, to experience this, I find the light not particularly blueish, but instead, the windows on the house give one the impression of extraordinarily yellowish illumination inside.

Which turns out to be ordinary neutral lighting, as soon as I have come inside again.

This is an example of chromatic adaptation. We usually experience the prevailing illumination as neutral, colourless. This makes it possible for the objects to display their proper colours.

There is a simple way of testing the chromatic difference between two illluminations. As, in this case, between indoor an outdoor light. On a table, near the window, you place a sheet of white paper. Fold a small piece of paper and put it on the table, like this.

The surface facing the window is strikingly blue; the opposite surface, facing the room and incandecent lamps, is yellow.

On the flat paper on the table there is a mixture of the two illuminations, with a soft gradient over the surface, from dominant indoor to dominant outdoor. When you have an abrupt change, such as at the sharp folding, creating two well-defined areas, bordering each other and facing different light sources, these areas will look as painted surfaces, seen in neutral light.

By the way, I cannot resist showing you a funny device, I once designed for an exhibition. You see a tower enclosing a room with five walls in different colours, and above it a dodekaeder, made of white paper. When it drops down into the room, its sides - as you see - get coloured with reflex light from the corresponding wall.

You can imagine, in practical life, in any situation, the prevalent illumination is a mixture of light coming from various directions, having been variously reflected and modified on its travel from the source - be it from painted walls and coloured transparent materials; be it from the sky or green foliage. The full mixture will mostly be experienced as colourless diffuse lighting.

 

OBJECT COLOUR CONSTANCY

Do the colours of objects change when seen in different qualities of illumination? Remarkably little, in fact. Normally we experience colour as a property of objects, by help of which we distinguish and identify them.

The perceptual constancy of object colours is generally acknowledged by the scientific community. Optical Society of America in their manifest "The Science of Color" declares: Color constancy is the substantial independence of object-color perceptions in the presence of changes in illumination or other viewing conditions.

The constancy can be studied as follows. You see here a montage of coloured papers. Several nyances of red and yellow, a couple of blue, a green, a brown, together with white and black as reference areas.

For yout information: In the actual demonstration I used a large format dia-positive photo of it (11 x16 cm) mounted in a black frame and laid on an OH-projector board. Thus the picture should be thought of as being illuminated from behind.

Look carefully now ... see what happens!

A transparent blue sheet is pushed over the picture, resulting in a number of mixture colours. With blue over yellow we get green. The blue colour of the sheet itself is seen over the white - it is almost identical with the blue paper to the right!
In general, blue areas get more saturated, whereas red areas become desaturated.

When the sheet covers the whole picture, this almost immediately returns to its original look - let be as seen in somewhat cooler light.

The sheet - now not seen as belonging to the picture any longer - has instead taken on the role of converting incandescent light (the OH-projector lamp) to simulated daylight.

And if we take it away - the illumination changes to slightly warmer character. But the picture remains the same.

The demonstration shows the effect of chromatic adaptation, compensating for changes in the quality of illumination, so as to keep perceptual object colours constant. Thus making it possible for us to recognize the composition of coloured areas.

Which quality of light shows us the true colours of the samples? It is perhaps meaningless to ask. Colours always look slightly diffferent in various light - this is a normal experience. Maybe we should ask: what illumination gives the most distinct and variegated display of the samples on the board? It is usually held that colour discrimination is best in ordinary daylight.

What governs chromatic adaptation? Is it that the brightest area should, as far as possible, take on the role of white? Or is it a common trend in the luminance shift of all areas making up the picture which guides the perceptual process? A kind of balancing act.

If one single area changes, we see it as a real physical change of the material at that specific area. But if all areas change -- in a typical way, our perceptual system is used to -- then we spontaneously apprehend it as a change in the quality of the overall illumination that has to be compensated for.

*

Let us consider a more intricate situation. A picture, composed of diverse colour areas, is hanging on a white wall, onto which a blue shadow falls. Can you behold the picture, despite this shadow? ...In an art gallery, probably yes. (And I admit, it is a bit exaggerated in the illustration below.)

You easily identify two white rectangles, one on the left side and one on the right side. On closer scrutiny, you find that the white stripe at the right side has the same tint of white as the background wall. So it seems that the blue shadow influences the colours in the picture.

Displaying the picture on black ground, makes things more intricate:

You don't any longer see the shadow, falling over it. And the question is: does it at all influence the look of the colour samples in the right half of the picture?

It must - you feel obliged to say... But don't be too sure..

 

EDWIN LAND'S RETINEX-THEORY

The great miracle of the eye is that it does not need what the physicists needs: it does not need uniform illumination in order to establish a lightness scale. (Land 1964, p.252)

Already in the 1960:s the ingenious Edwin Land, whom you may remember from my video on Two-Colour Projections, together with his collaborator John McCann made a series of experiments, the so called "mondrian demonstrations", similar to what I have just shown you.

They pointed out a simple fact: The colored area is never alone. On the contrary, in a natural situation there are lots of differently coloured surfaces making up the scenery.

They further pointed out that for the identification of surface colors, the borders between areas are more informative than the colored areas themselves. Which in fact solves our problem. Let me show you how.

Look at this simple picture. What do we see here? Well, we see a chequerboard pattern in black and white, illuminated from the right, making it slightly shaded on the left side.

The interesting thing is that if you compare a black square at the outmost right with a white square at the outmost left, they turn out to have exactly the same greytone. So what we are looking at on the screen is a pattern composed of greytones.

What is the objective truth here? That we see a regular chequerboard pattern in black and white, or a layout of grey tones?

Actually both. On the computer screen there is a layout of squares with various luminances, and this is what you are looking at now. It is the same as looking at a photo in black and white.
In a real, physical situation there is a difference between whether the luminance of a certain square is due to the reflectance coefficient of the paper or the local intensity of the illuminance. On a photo you cannot distinguish greys from shades. In order to do that you need further information, given by the situation. As in this case, by recognizing it as a checkerboard pattern in uneven illumination.
What I show you in this article and in the video are photograhic representations of real physical situations. In pictures there will often be ambiguities, but with material surface colour patterns in illuminated space, the invariance of object colours will easily be ascertained.

Let us for a moment analyze the example above from a physical point of view. We have a pattern built from squares with eighty percent and twelve percent reflectance respectively and below you see how it looks in homogeneous light.

Then you illuminate it with light falling from the right side from a close by source, giving an illuminance on the picture board like this

resulting in a luminance pattern like this, making a white square at the left reflect the same intensity of light as a black square at the right side, as we saw.

Now, suppose the eye registers the luminance jumps at the borders where areas meet. This will go up a factor of 7:1 (actually 0.80 / 0.12), then at next border down 1:7, next up 7:1, and so on, independently of the illumination at the place of that border. Multiply the five steps and you get 7. So the last square in the row willl be 7 times the first one, i.e white.
Thus, the registered pattern of surface reflectances would be as if measured in homogeneous light, shown in the first figure.

To get the true reflectances you have to normalize it, for instance in this example by assigning 0.80 to the brightest area.

Simultaneously, information about the gradient of the illumination is given by the successively darker surfaces (the mean luminance over the area) when going from right to left. So what we see, is a chequerboard pattern in black and white, illuminated from the right.

*

So far we have been dealing with plane pictures but real life takes place with three dimensional objects in space. Then the variability in the play of light and surface materials is even more intricate. Let me show you an example.

Trying to imitate chromatic adaptation, digital cameras adjust to the color temperature of a scene by flexible individual calibrations of the three channels R, G and B. Look at the following pictures of one and the same situation.

When directed towards the table the camera tends to make the white paper look white (lamplight 3400K) and then the computer screen (6500 K) looks blue. Directed more towards the screen, this may come to look grey, as it should in this case. But then the paper on the table is orange-yellow.

The point is now that for me, sitting there working at the screen, both the paper and the screen are seen as neutral, simultaneously. It seems that human vision is capable of handling two scenes with different illumination qualities, simultaneously present within the field of view.

This possibility was hinted at already by Edwin Land: "multiple color universes can coexist side by side or one within another" (Land 1959, p.120)

 

CONCLUSION

There is a continuously ongoing play between light and matter. What I observe depends on my intention, which governs how the informative factors are sorted out in the flux of light reaching my eyes. In practical life there will always be the task of identifying stable object colors under shifting illumination conditions. This is usually instinctively managed by the visual system with a sufficient degree of accuracy for practical purposes.

If your intention is instead to find joy in the wonderful ever shifting qualities of light and shadow, as the impressionistic painters did, you don't bother about letting objects have their "true", physical colours.
Light weaves everything together. Radiant bodies connect separate objects by illuminating them and thereby giving them force to illuminate each other and share colour with each other.

 

*

SUPPLEMENTARY NOTES

That the objects around us have familiar colours under ordinary illumination conditions is a self-evident experience. At least as long as we do not bother to observe and specify the actual colours with accuracy.

As a matter of fact the perceptual constancy of object colours turns out to be difficult to confirm by traditional experimental methods. The more you simplify the conditions, in order to have the visual situation under control, the less evidence you find for "colour constancy" - until you doubt that it is a special function of our visual system and not something you imagine. "We see what we are looking for", as the saying goes.

However, as among others Edwin Land pointed out, in a natural situation there are lots of objects and variously illuminated coloured surfaces in view, and this complexity of the situation seems to be a necessary condition for the phenomenon of "perceptual constancy" to appear. And with his demonstrations he managed to show that he was right so far.

The critically minded may still wonder: Is there really some such thing as the real colour of an object?

Well, at least from a physical point of view, yes. Listen to Optical Society of America: "Each real object has a so-called object color which is merely its capacity to modify the color of the light incident upon it. This capacity depends essentially upon the spectral reflectance of the surface and is a more or less unique and constant characteristic of the given object."

So the perceptual constancy of object colours could be due to the fact that what we perceive as the qualitative colour of an object surface somehow corresponds to its reflectance. This seems to have been Land's idea (Land 1974, p.28)

When I say corresponds to, I suggest that colour could be regarded as a measure applied to the spectral reflectance distribution, characterizing its elementary symmetry properties.

However, to determine the spectral reflectance by measurements is a difficult procedure. You have to measure the intensity of the reflected light and divide it with the intensity of infalling light, to get the reflectance factor. And this for a proper number of wavelengths within the visual range. One should also take into account different geometries: either diffuse illumination or various directions of infalling as well as reflected light. Meaning that the physical object colour is multifaceted - as is (maybe) also perceptual object colours.

What I, in my lecture, call chromatic adaptation is the visual systems inherent (and maybe through experience developed) way to reach that same goal, by adequately compensating for the influence of the local, momentary intensity and spectral composition of the prevailing illumination.

As soon as you enter a room, the first thing you instinctively detect is the character of illumination. A complex visual field is a favorable condition because of the numerous clues it yields.

Remember from my introduction (in the movie), when entering the house, there are lots of variously coloured surfaces, including white, and a situation with perceptually neutral lighting is quickly established, before one focusses one's interest on specific objects and their colours. This is the normal procedure, continuosly going on. In case of daylight through windows, the indoor light changes during the day, with passing clouds and the wandering sun.

*

Let us discuss the examples given in my video-lecture.

The folded piece of white paper. Here color constancy is violated. It looks very much like a blue surface colour. You definitely do not see it as "white" - not even as "white in blue light".

Note that it is because the illumination falling on the paper makes it look blue that we call it "blue light". We cannot see light itself. But if it makes a white paper (or snow field) look blue, we call it "blue light". As well as if we let a ray of this light fall onto a spot on the retina and you get the sensation "blue" - we call this light "blue", as did Newton in his famous investigation of the prismatic spectrum.

The perceptually established neutral illumination, in this case, is the additive mixture of the light from the window and from indoor lamps. And, of course, as soon as you move the folded pice of paper you soon recognize it as being white. And movement is permitted in the perceptual process, aiming at object colour.

In the mondrian-picture the colours before and after the blue filter are far from exactly the same -- I speak about "seen in slightly yellowish light" and "in slightly blueish light" respectively -- but they are sufficiently "the same" to make the picture as a whole look the same. And the chromatic adaptation effect is appreciable. Look at this little reminder.

To the left, the uppermost part of the yellow area is already covered by the blue filter, so that you see a green square there. To the right the whole picture is covered by the blue filter. All areas have changed, by being subtractively mixed with blue, except the green square, which has disappeared. It was a fake; didn't actually belong to the picture. The only area the luminance of which didn't change, when all the others did!

When both cases are shown together side by side, like this, the chromatic adaptation is weak. Neutral light favours the left one. Bu if you look at the right one separately, it gets a chance to brighten up, the white paper to look white, and even the others more like in the original picture, even if with less saturation, especially the yellow ones.

Finally, the checkerboard pattern, where the physically quite strong illumination change is almost unnoticeable. And the fact that the dark square on the right has exactly the same luminance as a white square on the left side. Here the constancy effect is strong - it is almost impossible to see the white squares at the left side as identical with the dark squares on the right side.

The effect could partly be due to Gestalt laws. As soon as the pattern is identified as "chequerboard" then it should be just black and white squares all over it. But Edwin Land has demonstrated perceptual constancy even for an arbitrary layout of greyscale tones.

Real space versus plane pictures

All examples you have looked at in this article (and in the video) are photographic pictures. This is a shortcoming of most research that has been done on visual constancy phenomena: the use of two-dimensional images. Then you cannot truly discriminate between illumination and object surface. Because light only exists in space, it is an aspect of space. Its role is to weave together the spatially localized objects, by being selectively reflected from their surfaces - including, of course, hitting the retina of the eye of an observer.

In a real situation you can separate illumination from surface by simply moving the object, or move yourself, as mentioned above. But this doesn't help, in front of a picture.

As psychologist James Gibson has pointed out, the perception of what a picture represents is a more complicated and demanding task than perception of a real spatial scenery. Where light belongs to the space in-between the objects.

I would even say: Investigation of perceptual constancies must be made with setups in ordinary three-dimensional space, in order to be authentic. Because in photos, drawings, or on an electronic screen, spatial depth is illusory and in addition there appear lateral interaction effects, you don't get in spatial situations. And these can lead you astray. Such as simultaneous contrast, afterimages, transparency, and the like. (I once, for an exhibition, made a computer program where you can study 12 different such effects, very amusing!)

*

The skilful and learned Ralph Evans, researcher at Kodak labs, in his last book introduced an important distinction for the study of colour perception. He says:
"... we unconsciously adopt from physics and mechanics the unitary nature of the outcome of a known system of causes. In doing this we err .. the error comes in the belief that what is seen must be a single entity. This is demonstrably false in perception."
"There is nothing metaphysical about the statement that when an observer looks at a natural scene he sees the illumination as separate from the objects being illuminated. Under normal conditions an observer looks at situations, not at stimuli as such."

I would express it like this: Normally we do not see the retinal image; we see by help of the retinal image, and by help of the light emanating from the observed object surface.

Evans then goes on introducing three "frames of reference" for perception: illumination, object and stimulus. And says: "To the naive observer the object frame of reference may be the only one of which he is ever aware and then, not as a separate phenomenon, but as the whole of his visual experience."

In practice, you become aware of the illumination as such when adaptation is not perfect, so that you experience difficulty in recognizing objects. But what about the "stimulus" frame?
That's when we become aware of the subjective side. Normally, when moving in the room, I experience it as a stable situation; objects having there constant shape, sizes and positions. If I experience it from my subjective point of view -- as if I were standing still -- then parts of the furniture seem to move an get deformed nightmarishly.

More difficult is when you are standing perfectly still and try to observe the scenery as if it where a plane picture in front of you.
This is the attitude of the artist who wants to find out how to present the situation ( the motif) in the form of a pictorial representation that - in the best of cases - will be more real and appealing than the objective reality itself.
It is then necessary for him to carefully observe how the quality of the illumination and surface reflectance are weaved together in the projected image presented as stimulus to the eye.

He must try to create an illusion of the presence of light in the scene of the painting. Even if real light only falls onto it! So what the painter trains himself to do is avoiding the illumination and object modes of reference and thus avoiding the perceptual constancies - which are in practical life so important. Just in order to be able to create an illusion of objective reality instead.

That is a technical aspect. But more interesting: Paradoxically, he may manage to create an illusion that is more convincing and tell the beholder more about reality than physical reality itself, which has a curious tendency to avoid our efforts to grasp it directly ... Hence the importance of art, artistic interpretation of the world, nature, sceneries.

 

LITTERATURE

David Katz: "The World of Colour" (Kegan Paul 1935) quote p.2 / orig "Der Aufbau der Farbwelt", Leipzig 1930.

Edwin H. Land: "Color Vision and the Natural Image", Proc. Natl. Acad.Sci. 45 (1) Jan 1959
Edwin H. Land: "The Retinex", American Scientist 52 (2) June 1964
Edwin H. Land and John J. McCann: "Lightness and Retinex Theory" Journal of the Optical Society of America, Vol 61 No 1, p.1-11 1971
Edwin H. Land: The Retinex Theory of Colour Vision" Proceedings of the Royal Institution of Great Britain Vol 47 p.23-58 1974 Mondrian picture, shown in the video, from this article.

Ralph H. Evans: The Perception of Colour 1974, therein spec. Frames of Reference p. 191-199 and Adaptation p 224-229.

James J. Gibson: "The Ecological Approach to Visual Perception" 1979. Therein especially chapter 15, "Pictures and Visual Awareness".

Hans Jürgen Scheurle: Überwindung der Subjekt-Objekt-Spaltug in der Sinneslehre 1977, 1984 Georg Thime Verlag, Stuttgart.
cit. "Wer aber nur Objekte sieht, ohne zu wissen das er sieht, kennt auch die Objekte im Grunde nicht."

Martin Buber: Der Mensch und sein Gebild (1955) and Urdistanz und Beziehung (1951).

Optical Society of America: The Science of Color. Copyright 1963; Chapter 5, color constancy, p.153-158.

For further reading I recommend:
(ed) Alex Byrne and David Hilbert: Readings on Color, Vol 2 The Science of Color. (1997)
Therein you find 3 essays on color constancy by E.Land, Brian Wandell, and the well-known pair Dorothea Jameson & Leo Huvich (p.141-198) and also Roger N. Shepard´s thoughtful essay: The Perceptual Organization of Colors - An Adaptation to Regularities of the Terrestrial world? (p. 311-356). See especially p. 318-325

 

My own extensive analysis of E. H. Land's two-colour projections can be found here