Daylight color temperature and visible color

[Brandt Wilson]

Hi,

This is a basic question that's been bugging me. With an HMI being so blue on the kelvin spectrum, why does it (and daylight) look gold/orange (warm) on the subject? Do our eyes want to see a warm cast to the dominant light source and adjust accordingly?

Thanks!

[Marcus Marchesseault]

You guessed it. Our eyes/brain "white-balance". Many of the things that film and video people spend their time on are techniques to simulate tasks that our eyes and brain do automatically. There are many instances of this and I probably only know a few. Here are some:

Our brains "stabilize" the jumpy image that is created by a bunch of rickety bones ambling along on two legs when most of nature seems to agree that four legs are better. Do you constantly notice that your head moves up and down while you walk? Probably not as your brain has done some stabilization.

Our eyes rotate in their sockets to remain level to the horizon. This is another one of the reasons why a jumpy camera looks bad. I also think that "Dutch" angles must be greater than this degree of rotation (I'm guessing 15 degrees) to look good. We never see images in our brain slightly off-level. Things are either level or very off-level. This is the reason for the leveling bubbles on tripods.

Our eyes have a shallow depth of field. This is why it is okay for a large part of a scene to be out of focus. As long as the main subject is in focus, it works for our brains. This is why people are getting interested in 35mm adxapters to allow shallow depth of field. In the real world, we move our eyes and re-focus almost instantly to simulate everything being in focus. On the screen, our eyes are drawn to the subject in focus. This allows the cinematographer to use natural tendencies to bring attention to a specific subject.

Our eyes adjust rapidly to changing light levels. This is why we must keep exposure under constant control while filming.

We perceive dim light as soft light. Candles are actually a harsh point source, but we thing they are soft because they are dim. The moon is another instance of this effect. We don't perceive moonlight as the point source that it really is. Because of this, you should use diffusion on studio lights you are using to simulate candles or the moon.

Our eyes move very rapidly. Except in extreme circumstances, we do not perceive vibration to our heads visually. If our bodies are jostled, we keep our vision fairly stable. This is similar to the effect of our brains/eyes stabilizing walking motion only working with higher frequency motion.

Our eyes are great with color vision compared to most creatures, but we still see black and white detail the best. I think we have more rods than cones in our eyes. This is why video compression can throw out much of the color detail as long as it preserves the luminance (black and white/brightness) information.

Although our eyes have no zoom function, our brains can focus our attention on objects at greatly varying distances. This is why cameras have lenses of multiple focal lengths. Using a long lens to limit the field of view is akin to our brain concentrating on just a single important subject. A movie crew can put the camera anywhere they want. Different focal lengths are not necessarily used to zoom in, but rather are used to change the shot composition in the way our brains vary what we are concentrating upon.

I'm sure there are many other techniques used to simulate what our eyes and brains do automatically. In fact, I would love to hear about them. If there is a book on the subject, I would love to read that book. Any ideas?

[Glenn Chan]

Sunlight will sometimes appear warm/orange-y because it is relative to other light. The light from the sky is very blue+violet, and appears much blue-er than sunlight.

Your eyes/brain kind of have an auto-white balance, so you will see the sunlight pretty close to white while the shadows will be blue-ish.
However, it's not so simple as auto-white balance:

A- Your eye/brain seems to add image processing to what you see (we don't know if it's the eye or brain that applies this processing). It is very good in achieving lightness constancy. Explanation:

Luminance is the light intensity as measured by an instrument.
Reflectance is how much light a material reflects. For this discussion, ignore specular reflections and assume diffuse reflections.
Illuminance is how much light is striking a surface.

Luminance = Illuminance * Reflectance

If you take an image with a camera, you only have one part of the equation (luminance). However, it's useful to know the reflectance of an object as it helps to identify them. Reflectance is a property that doesn't change (ignoring diffraction [i.e. on CDs] and specular reflections), whereas illuminance and luminance are always variable.

Since you only know one variable, reflectance can only be guessed. Hence the lightness constancy *problem*. Various heuristics / assumptions can improve the accuracy of your guesses. One of them is this: In a real-world scene, there are probably a range of objects from little to very high reflectance. We can assume that the highest luminance in a scene corresponds to an object with close to 100% reflectance. So based on this assumption, we can figure out what the illuminance is. If illuminance is the same for one scene, then we can determine the reflectances of everything else.

Of course, in real-world scenes illuminance is NOT constant due to shadows. Our vision tries to compensate for this, which explains why a grey rectangle superimposed on a gradient looks like it has a reverse gradient in it.

Lightness is the *perceived* reflectance, as guessed by our eye/brain.

B- Our eyes also seem to achieve color constancy. This is figuring out the color of an object without knowing the color of the incident light. i.e. Sunlight is very orange compared to sky light.

C- Our eye's white balance seems to be affected by the following factors:
A- The brightest object in a scene is nearly always assumed to be the color of the incident light.
B- "saturation" (probably the incorrect term) also affects thing. Highly "saturated" colors are probably not white (reflecting R, G, and B about equally). So during golden hour, your eyes may see the sunlight as golden (instead of white). Also, your eyes typically will not fully white balance to tungsten light (3200K) or low-temperature incandescents (2800K and lower; 'nite lights' for putting kids to sleep are a good example).
C- Angle seems to matter. Your eyes will have a tendency to white balance to whatever computer/television monitor its looking at when faced with a multiple-monitor scenario.
D- In shadows, our eyes do not assume A. Which makes sense, because shadows are blue-er than everything else.
E- Consumer TVs typically have a very blue white point. I may be wrong here, but it seems that our eyes adapt really quickly to very blue color temperatures (possibly because shadows are typically very blue?).

At this state of adaptation, reds appear less saturated than they would at D65/D6500. To compensate for this, many consumer TVs will over-saturate the reds.

Also see:
Google and learn about "lightness constancy", then "color constancy".
Go to http://tig.colorist.org/wiki2/index....ception_Quirks and scroll down to the link on the Edward A. Adelson paper. I think it's password-protected, but you can grab it off the Google cache I think.

Some other information at:
http://www.brad.ac.uk/acad/lifesci/o...pvp1/index.htm
(I think I disagree with some of the information there, but nevertheless the information there is good.)

[Mike Teutsch]

Great response guys and enjoyed reading them. Fantastic thing the human brain!

Thanks very much---Mike

[Marcus Marchesseault]

Okay, have you ever wondered why adding Red, a low frequency color, to Blue, a high frequency color, the outcoming color is not a mid-frequency color? What do you get when you combine Red and Blue? You will get Magenta (Purple). Where is purple on the color frequency list? It's not there! There is Violet, which looks purple, but it is a higher frequency than blue.

Go here for the answer:

http://www.marktaw.com/design/ColorTheorya.html

The gist of it is that our eyes have a complex range of frequencies that each Red, Green, and Blue cone receive. Magenta (Purple) is a color that our brains interpret for us when there is light with a lack of Green. Purple is all in our heads.

[Brandt Wilson]

Thanks guys, lots of unexpected information here.

Okay, here's part 2 of the question, starting from a real-world example:

Five years ago I directed my first project. The DP shot a mix of incandescent softboxes and daylight, so the color balancing went REALLY haywire when I went to color balance in post. A couple things I found were that even if we shot under all incancescent or all natural sunlight, I couldn't match them with a white balance. I don't recall the specific differences, but the tonality was decidedly different. It took a lot of addition work in the highlights and shadows to get them to align. Does this have to do with how lights of different base color reflect differently off surfaces of varying colors?

[Glenn Chan]

You can google metamerism. Maybe that is it?

http://en.wikipedia.org/wiki/Metamerism_(color)

Good real life examples would be:
Street lights versus everything else. Sodium vapour lighting emits light only at specific frequencies, instead of over a range of frequencies. Also, it doesn't emit a blue component of light.

Fluorescents are another example.

Day versus night is another (i.e. nighttime in the city, which is not so dark you no longer see color).

[Doug Boze]

Originally Posted by Marcus Marchesseault

You guessed it. Many of the things that film and video people spend their time on are techniques to simulate tasks that our eyes and brain do automatically.

You've nailed it, Marcus. We see, hear, touch, smell, and taste with the same organ: our brain. If you hold up a piece of paper (yes, white typing or printing paper) and ask all present, "What is the color of thine paper?" And the response of the faithful is, "White!"

Conditioning, mainly. Even if the question is asked while a red, or blue, or green lamp is shone upon said paper, the answer is the same, and also correct. Objective vs. subjective.

Our brains say the paper must be white, hence, it [i]is[/is] white. Cameras, both film and CCD, are absolutely objective, hence, "what you see is what you get. That page is sorta green, or blue, or something. Why are you complaining?" we complain because of the objectivity. Sort of like facists bitching that their crimes have become public.

But we can work with perceptions. Skew things one way, and it looks good. Skew them another, and they seem bad. Which are they? The truth is somewhere betwixt and between.

At which point does perceived reality become art, or vice versa? That is the question. Are you creating or setting mood, or are you representing fact? That, too, is the question.

Or a question.

Or none at all.

Think about it.

[Doug Boze]

The human eye, or aye, for those of you in their cups, is purely subjective. We can only see three (3) colors:

RED

GREEN

BLUE

...and any other color is pure conjecture. It is with our brains (those oh-so-unreliable-difference-engines) that we perceive. Hence hallucinations.

Everything we see is merely a mass hallucination. Get over it. It has worked before, and it has worked marvelously for Faux News.

[Marcus Marchesseault]

Check out that link I posted. I found it fascinating. Apparently, our eye's color cones have a great range of frequencies. The red and green cones have almost the same range and sensitivity profile, only they are shifted in frequency. The blue cones are less sensitive and range from ultra-violet down to cyan and a bit of green. There is no color frequency for purple. Ultra-violet looks purple, but purple pigment is not ultra-violet. It is just not-green.

Another interesting thing is that all of the color cones have a common frequency and sensitivity somewhere in the cyan range. All the cones see a cyan color with the same strength. Weird.

Also, up is down. No, really. All lenses, including our eyes, invert images.

As for part 2 of the question: "A couple things I found were that even if we shot under all incancescent or all natural sunlight, I couldn't match them with a white balance."

White balance can be tricky. It is best to do it accurately for every scene change, because the light on your subject is effected by so many reflected surfaces and changing light conditions of the source (clouds, etc.). If you don't care as much about perfect accuracy and would rather have consistency, set your camera to daylight (5500K) or incandescent (3200K) and leave it there. Of course, choose the appropriate fixed balance color for your conditions. I assure you, no subjective brain will like a scene shot indoors with incandescent light on a camera set to daylight balance.

[Graham Bernard]

. .and of course because filmmakers "know" how our eye and brain work they break the rules to instill another effect. The simple one is the 25fps - we can't register above 25 - but I bet there will be some insects - I'm guessing here - if they were subjected to a flickering 25fps, who'd stomp out of the cinema demanding their money back!

Know and understand the "Human Rules" and yah can break 'em! - Just don't expect spiders to offer you an Oscar though.

Grazie

[Marcus Marchesseault]

Actually, about 25 is the minimum we can interpret as smooth motion. 24fps film was set that way as the minimum speed allowable. Anything worse than that, and it looks too stuttered. For smooth motion that does not have motion blur, at least 30fps is the minimum with at least 40 being preferable. 40fps is the speed that Imax runs their film for smoother motion than standard 24fps 35mm film.

As far as the actual high speed that the human eye can detect, it is at least 100fps for some people. Check out the refresh rate of your monitor and you will probably see something like 85Hz. Now, in a dark room, wave your hand back-and-forth in front of a CRT monitor and you will be able to see the strobing created by the fast refreshing CRT. Instead of a blurred skin-tone image of a roughly hand-shaped object, you will see many "copies" of your hand frozen in time. The ability to detect this strobing is your eye's ability to see MUCH faster than 25fps. Fortunately, for us and the motion picture industry, our eyes don't work with a fixed frame rate and interpret motion blur as the correct image to receive from fast-moving objects. Our eyes are constantly exposing so we have information coming in seamlessly.

[Richard Alvarez]

24fps was settled on for purposes of sound reproduction, not 'flickerless' imaging. Prior to 24fps, it was often 20, or 18 depending on the country of origin and system in use. Plenty of super 8 film shot at 18 frames per second, looks just fine projected on the screen.

[Marcus Marchesseault]

"looks just fine projected on the screen." Well, we'll just have to agree to disagree on that. I think 24fps looks horrible in many shots. I am personally totally removed from the motion picture "experience" when a sweeping pan or fast dolly shot creates a stuttering foreground. Yes, most filmmakers work around the 24fps limitation quite successfully, but it is still a serious limitation in my book.

I thought of another thing that our eyes/brain does that is often faked in movies. We tend to interpret the moon near the horizon as looking larger. Have you ever looked out at the moon when it is near the horizon and noticed how large it looks? It doesn't always happen, but I have noticed people commenting on "look how big the moon is tonight" several times in my life. The reality of the situation is that it is an optical illusion created by the moon's proximity to ground objects in our field of view. The moon (and sun) never changes size significantly in relation to our eye's field of view. It just seems more prominent shining through trees or a city skyline. The sun looks larger as it goes down over the horizon. You can check this illusion yourself (I did and still had a hard time believing it). When the moon looks really huge some evening, hold something out at arms length to compare it's relative size to that fixed object. I think I used my thumbnail with one eye closed. If you are really determined, this could be a ruler or tape measure. Then, wait a few hours for the moon to move high in the sky and re-measure. You will find that it hasn't changed size, but it looks smaller.

Edit: I forgot to say that the moon and sun go down in color temperature as they get closer to the horizon. The moon/sun light passing through the atmosphere at a greater angle has more of it's blue frequencies scattered or absorbed. If you want to simulate this with lighting (since this is a lighting forum), have a gold reflector or light gelled down in temperature hitting the actors from a horizontal angle. The more blue light from above and all around contrasted by a lowered temperature light from the side can make it look like sunset. A diffusion fram or "silk" above will cut down on the harsh shadows. I seem to recall reading on this forum that "CTS" (Color Temperature Straw?) gels may be favored for this application. Apparently a temperature-lowering gel with a bit more yellow will give that golden sunset look a bit better than the standard CTO (Color Temperature Orange). I personally have had some success adding a bit of minusgreen along with the CTO.

[Brandt Wilson]

So, do I have this right?

For a daylight scene representing a clear sky with sunlight on the subject:
Fill: 9000k
Key: 5600K

White balance from a white target in the fill lit area.

Will this provide a natural sunlit look?

Thanks,
Brandt