PDA

View Full Version : color shifts in shadows (reboot)


drollere
04-30-2005, 08:59 PM
the issue of the color shifts in shadows ought to be discussed in terms of objective color measurements or specific color observations. then, at least, there is a common framework for interpretation.

COLOR MEASUREMENTS

i've reproduced below patrick1's demonstration of the pencil shadows -- the one i requested, on white paper -- with a photoshop analysis.

first, i've patched out the THREE relevant color samples: the shadow produced by sky light, the shadow produced by the light bulb, and the unshadowed color of the white paper. i've also marked out the exact points where i took the samples (3 pixels below the white dots along the white line), using 5x5 pixel sampling. using photoshop, you should get identical results.

to settle the dispute about what "color" (hue) the shadows are, i boosted each sample up to its maximum chroma. this makes things unambiguous.

i should mention that, for the paper color, i could get almost any hue i wanted -- red, orange, yellow, green, cyan, blue, violet -- if i sampled around long enough. (it's a minor point: the color is so close to a perfect neutral that random error shifts it around.)

below the pencil photo, i've attached a simple hue/chroma analysis of these shadows, using the CIELAB values reported by photoshop. these obviously ignore any contextual effects.

the fundamental physical framework for analyzing these kinds of shadows is:

(1) "white" = light 1 + light 2
(2) shadow 1 = light 2
(3) shadow 2 = light 1

that is, the eye adapts chromatically to the average of all the illumination in the scene, direct and reflected, strong and weak, and judges colors against this 'white' standard (or its gray equivalent).

against this standard, the weak (light bulb) shadow illuminated by the strong sky is obviously "cool" (close to cyan or cerulean), while the strong (sky) shadow illuminated by the weak light bulb is obviously "warm" (orange tending to scarlet). the axis of the shift is roughly parallel to the warm/cool dimension on most color wheels.

this is in fact the general effect of simultaneous shadow contrasts in natural lighting, produced by direct sunlight and indirect skylight, first reported by 18th century naturalists and amplified poetically by goethe, among others. it also appears in *simulated* shadow contrasts produced by an incandescent light source (tungsten bulb, candle, any other "black body" illuminant) and sky light (as in patrick1's photo).

the color shift is sometimes thought to be universal, and has been codified as a general rule: "warm light = cool shadows, cool light = warm shadows". this *general* rule is false, as i will show in a moment.

however the natural contrast arises as part of the visual mechanism that adapts our implicit white point or "sense of white" to changes in the natural illumination, specifically the reddening of the sun's light near the horizon and the bluing of daylight by atmospheric scattering. we live with it and live in it, even on overcast days.

this red/blue shift is exactly paralleled by a planck blackbody source, which is why photographers can usefully refer to lighting in terms of its blackbody temperature, and why our visual system has a "prebuilt" sensitivity and prompt response to this temperature change. we adapt to the warm/cool color of illumination about as quickly as we do to the intensity of the illumination, from dim light to bright. (that is, we easily adapt to these light changes faster than changes in sun or sky can produce them.)

if the visual white standard is the sum of two separate light sources, then obviously it will be between them on a hue/chroma plot, in the same way the mixture of two paints is in between the original two paints. the CIELAB plot shows this, to an approximation.

in comparison to this average, eye adapted "white" value, the light from the bulb will appear warmer wherever it illuminates a white surface by itself, and the sky color will in the same situation appear cooler.

the color depends on the color of the light, on the eye's mechanism for adapting to the color in the light, and on the eye's adaptation level, along with many other things. but it is a robust effect.

note that the color of natural shadows is not just dependent on the intensity of the source casting light *into* the shadow but is also dependent on the intensity of the source *creating* the shadow: the chroma of the orange sky shadow is higher, even though the light bulb is a weaker light source (its blue shadow is fainter). darker shadows produce clearer color contrasts.

COLOR PERCEPTIONS

i posted a summary of colored shadow contrasts in "shadow ii" that got not a single comment, not one, amid the flame fest. so i'm reposting it here.

i used four colored flood lamp lights purchased at a hardware store. these are nominally red, yellow, green and "blue violet" (near UV) lamps, though i can't vouch for their emission spectra.

i observed the lamps in all combinations, and each combined with a white (tungsten) light, in the two illuminant shadow situation, and observed these effects:

(a) blue violet + green = "white"
shadow green = blue violet (!)
shadow blue violet = green (!)

(b) red + yellow = "white"
shadow red = green
shadow yellow = red (!)

(c) red + white = "white"
shadow white = white (!)
shadow red = green

(d) green + white = "white"
shadow white = gray
shadow green = red

(e) yellow + white = "white"
shadow white = gray
shadow yellow = blue violet

(f) blue violet + white = "white"
shadow white = gray
shadow blue violet = gray (!)

before i proceed: THIS IS A PAINFULLY SIMPLE DEMONSTRATION to do for yourself, if you have a windowless room and a local hardware store and a smidgen of adult initiative. so please do not comment on these effects until YOU HAVE OBSERVED THEM FOR YOURSELF!!!

the first point is that the "warm light = cool shadow" rule is obviously false in every situation. this is obvious in the case of the blue and green lights: the blue should produce a yellow shadow, and the green a red shadow! neither occurs.

another glaring exception is that a blue violet light does not produce a "yellow" shadow color, but a gray. i am still not sure of the cause, but it is either that (1) an unsaturated yellow resembles a raw umber, that is, it is rendered as a dark yellow but dark yellows just look gray, or (2) the near UV light contributes almost nothing to the total level of illumination, and this is necessary for a color shift to occur. note the suggestion (1) that these shadow colors are actually visually interpreted as *surface* colors, as there is no such thing as a grayed or umber color in aperature or light colors.

the second point is that the strongest effects seemed to me to be produced by the red-green contrast, specifically by the red light. these seem to overwhelm other effects, as shown for example by the red + yellow.

the third point is that there are obviously several things going on at once. when both lights are cool, shadow shine occurs (blue violet + green). when one light is white, a vivid color contrast occurs, but not for a blue violet light.

the basic context here is similar to that for color constancy or discounting the illuminant. under situations of natural lighting or something close to it, the adaptive mechanism works in intuitive ways. under unnatural situations the adaptive mechanism is more fragile -- much as color constancy breaks down under sodium vapor illumination.

again, please do yourself the favor of looking at these effects for yourself. dogma does not substitute for the facts.

Einion
05-01-2005, 11:45 PM
Thanks very much for the further analysis of Patrick's twin lightsources example, it is most appreciated.

to settle the dispute about what "color" (hue) the shadows are, i boosted each sample up to its maximum chroma. this makes things unambiguous.
I thought it was unambiguous in context, some people might have chosen not to examine the picture too closely and have it clash with a dearly-held view but being as objective as I know how I thought the point was made clearly.

i should mention that, for the paper color, i could get almost any hue i wanted -- red, orange, yellow, green, cyan, blue, violet -- if i sampled around long enough. (it's a minor point: the color is so close to a perfect neutral that random error shifts it around.)
The fact that it's so close to perfectly neutral is significant yes, but even without a neutral background the nature of the illusion was still worth considering and examining for the lesson it provided about the relative colour judgement I thought.

the color shift is sometimes thought to be universal, and has been codified as a general rule: "warm light = cool shadows, cool light = warm shadows". this *general* rule is false, as i will show in a moment.
Thank you, something I have argued any number of times in the past.

note that the color of natural shadows is not just dependent on the intensity of the source casting light *into* the shadow but is also dependent on the intensity of the source *creating* the shadow...
Thanks for this observation, I'm sure it will come in most useful.

i posted a summary of colored shadow contrasts in "shadow ii" that got not a single comment, not one, amid the flame fest. so i'm reposting it here.
Actually it did get a comment, I made a point of it:
Thanks for your post Bruce, lots of interesting information to consider. I'm very fond of examples that show 'rules' to be anything but and you've provided a couple of good ones here.
I didn't make any further comment for two reasons that I can think of, one, I didn't consider it necessary to debate the veracity of anything you'd stated (complex reasons for this that I don't have time to get into, sleep awaits) and two, because of the difficulty in extrapolating from your examples to the subject at hand. And after a while one tends to forget with other things going on... BTW, you must have seen a lot worse in exchanges online so to describe the rest of the thread as a flame fest is to unfairly characterise it. Given that you state later that you don't welcome comment from anyone who hasn't observed them for themselves perhaps it's best that nobody did?!

so please do not comment on these effects until YOU HAVE OBSERVED THEM FOR YOURSELF!!!
I thought emotions were absent from your posts?

Einion

Patrick1
05-04-2005, 11:20 AM
the eye adapts chromatically to the average of all the illumination in the scene, direct and reflected, strong and weak, and judges colors against this 'white' standard (or its gray equivalent).
Is this what 'adaptive state' means? Similar to when you wear colored sunglasses, then after a few minutes, their color is no longer very noticeable and colors start to look closer to 'normal' again.

Or a brightness adaptation: like the way that when you've been in a dark room for a long time (like the movies) and then go outside on a bright day, everything looks way too bright. Or the opposite: going skiing on a bright clear day without sunglasses, then you go inside and everything looks way too dark.

In each case, after a while your eyes 'adapt' and things look more 'normal' again. I always assumed that this is what adaptive state means...but would like to make sure.

If this is what it means, it's similar to music; play a piece of music on an instrument. Then play it the same, but transposed up or down by a certain amount...say a half step or whole step. At first it sounds wierd but after a while, it'll sound normal again.

drollere
05-23-2005, 06:08 PM
i've conducted a more comprehensive test of shadow contrasts and have come up with results that are straightforward and trustworthy. i am delighted with the results.

i don't repudiate the observations posted above for colored lights, but it is clear to me that the "color" of the light masks the actual bandwidth or composition of the light, and while this won't affect the color of the light seen directly, it will affect the *mixture* of the light reflected from a colored surface -- which is how we perceive shadows in all cases -- so it can affect the color of the shadow.

so general rules will be very difficult -- unless we very familiar with the illuminant, which is true for the sun and most incandescent artificial lights.

CHROMATIC ADAPTATION

first, to patrick1's question, let me explain how chromatic adaptation works. the dominant effect is in the shift of natural light from "blue" noon sunlight to "red" sunset light, and this has been extensively studied as the contrast between a simulated "daylight" illuminant and an incandescent (yellowish red) light source. when the brain is completely adapted to normal daylight adaptation the illuminant (*) will appear to have roughly the same color to the brain as white surface (W), and therefore the surface will appear "white".

light: b----------(*)----------y
brain: B----------W----------Y

as the sun goes down, the color of the illumination goes toward yellow, and this has two effects: it makes a light yellow colored surface appear to be "white," and it makes the actual white surface appear to be "blue".

---->b-----!----(*)----------y
B----------W----!-----Y

(the brain adapts to this up to a point, after which we notice the "color" of light, for example around sunset.) now, the same effect occurs if the mind adapts to a colored light, such as a yellow light, in terms of shadows. shadows are perceived as darkened in a neutral direction, but this "neutral" now is shifted into the blue. hence a yellow light produces a deep blue colored shadow.

OBSERVING SHADOWS

i had tried using colored spot lamps, the kind you can get in any hardware store, to look at shadow contrast effects, but the results were hard to summarize. this made me doubt the lamps and decide to use kodak wratten filters to use instead.

i used two white (incandescent) light projectors set up to create overlapping shadows of a large cylinder onto a full sheet of white bristol paper taped to a wall. each projector was used to project white, red, yellow, green, cyan, blue violet and magenta light in 21 (all possible) combinations. i recorded the shadow color with a photoshop file displayed on a nearby apple cinema display monitor; the file contained two large color rectangles displayed on a black background that i adjusted with hue/saturation/lightness controls to visually match each shadow. i then wrote down the color specs. i also described each color with paint or color names.

desipte the fact that many shadow colors were out of the computer display gamut (as well they might be, being imaginary), and greens were very difficult to match due to monitor limitations, this method turned out to give really reliable (repeatable) and apparently accurate results.

COLOR + WHITE

in shadow combinations of color and white, the shadow cast by the colored light takes on the light's *visual* complementary color. (the visual complement is the *color opposite* the light on a munsell or cielab hue circle.) i noticed that this shift was slightly greater for the green, cyan and yellow lights, and slightly less for the red, magenta and purple, i think this is an artifact of the fact that the filters passed unequal quantities of light.

the general rule is: the shadow of a light with a hue angle (location on the color circle, in degrees) of H will create a complementary shadow color of H+180:

H shadow1 = C light1 = H light1 + 180

COLOR + COLOR

the results here turned out to be amazingly consistent. the filters allow pairings of colors that are separated by about 180, 120 and 60 degrees on a visual color circle. there is a simple mathematical rule to predict the color of the shadow:

H shadow = (C light + H light2)/2

if the lights are visual complements, such as red (hue angle 35) and cyan (hue angle 210), then:

H shadow red = ((35+180) + 210)/2 = 425/2 = 215 (cyan)
H shadow cyan = ((210+180) [-360] + 35)/2 = 65/2 = 32 (red)

[you have to subtract a whole circle, because a complement of 390 just means a hue 30 degrees past 0.) in other words, if the lights are complementary so will be their mutual shadows. for lights in a triadic relationship of about 120 difference, such as red and green (hue angle 150):

H shadow red = ((35+180) + 150)/2 = 365/2 = 182.5 (turquoise)
H shadow green = ((150+180) + 35 [+360])/2 = 725/2 = 362.5 = 2.5 (carmine)

and for lights in a tertiary relationship of about 60 degrees difference, such as red and yellow (hue angle 90):

H shadow red = ((35+180) + 90)/2 = 305/2 = 152.5 (middle green)
H shadow yellow = ((90+180) + 35 [+360])/2 = 665/2 = 332.5 (magenta)

[again, you have to add 360 to indicate that the mixture is in a clockwise rather than counterclockwise direction on the hue circle. a nuisance.]

the general mechanism is this: the colored light produces its own complementary color in its shadow, but this shadow is illuminated by a second colored light. the mind "mixes" or averages these two colors just as if they were both physical lights.

i have to mention a neat case: the mixture of red and magenta. my notes were that the shadows for these lights were blue violet and sap green. i couldn't figure this out. in fact, it is entirely predictable. the complement of red is cyan, which mixes with the magenta to produce the blue violet. the complement of the magenta is green, which mixes with the red to make yellow -- but this yellow is dull, and therefore looks like a green gold.

so you have the case of *two* "warm" lights, and they make a pair of "cool" shadows!

FriendCarol
05-29-2005, 12:17 PM
Apologies; I have no windowless room (other than a very small bathroom), and although I have a friend's projector still in its box somewhere in the house, I have no filters, and no access to a hardware store (and no second projector). So I cannot repeat your experiment, but I do have a few questions. Okay, only one question:

The complementary shadow seen when white light illumines the shadow cast by one filtered light -- that result is due to chromatic adaptation?

If not, where on earth is that second color coming from?

Apart from that, it seems straightforward. :p Thanks.