• Please use real names.

    Greetings to all who have registered to OPF and those guests taking a look around. Please use real names. Registrations with fictitious names will not be processed. REAL NAMES ONLY will be processed

    Firstname Lastname

    Register

    We are a courteous and supportive community. No need to hide behind an alia. If you have a genuine need for privacy/secrecy then let me know!
  • Welcome to the new site. Here's a thread about the update where you can post your feedback, ask questions or spot those nasty bugs!

sRGB - the white point and recommended viewing environment

Doug Kerr

Well-known member
The Story about the why and how of white balance color correction basically says at the end, "And that's why we want the illumination in the place where we view the image to have the same chromaticity as the white point of our color space." (That a teaser; you will be subjected to The Story - in an abridged form - shortly.)

But for the sRGB color space, the white point chromaticity is that of standard illuminant D65, while the recommended illumination of the viewing room has the chromaticity of standard illuminant D50. How come?

Many discussions of the sRGB color space address this with facile explanations, but upon closer examination they are like "Although the Moon is smaller than the Earth, it is farther away." But this is what I make of it.

I will start with a (fairly-concise) rendering of The Story of white balance color correction.

Imagine that we have a piece of note paper whose reflectance is uniform over the range of visible wavelengths. It certainly seems to deserve the description "white" paper. For such paper, the chromatically of the light reflected from the paper will be the same as the chromaticity of the illumination on the paper.

If we first observe the piece of paper at, say, the dining room table under incandescent lighting (remember incandescent lighting?), and later observe it at our patio table under sunlight, the color of the light reflected from it will differ appreciably. Yet our perceptual system in each cases probably judges the "color" of the paper to be the same.

This happens because of the extraordinary ability of the human visual system to outguess what is going on here. From the color of the light reflected from various surrounding objects (many of them familiar), the visual system in effect deduces what is the chromaticity of the illumination, and "backs that out" of the color of the light that strikes the retina to determine the "reflective color" of the piece of paper.

Now we move from direct visual observation of that piece of a paper to a photograph of the scene it is in. Suppose we first do that in the "outdoor" setting. And suppose that our end-to-end image chain results in the color of the "piece of paper" on the viewing screen having exactly the same chromaticity as the light reflected from the paper had. But further suppose that the image is viewed on-screen in a room illuminated by, for example, incandescent lighting.

The visual system sets up to deduce the reflective color of that item in the image assuming (incorrectly) that it is really right there, being illuminated by the light in the viewing room. The result will be is that to the viewer the paper will look "bluish", not white.

Outwitting this phenomenon is the purpose of white balance color correction.

[To be continued]
 

Doug Kerr

Well-known member
[Part 2]

The white point of a color space is (somewhat simplistically) defined as the chromaticity that is represented by some certain "handy" relationship among the coordinates. (In RGB-family color spaces, that is R=G=B.)

At the receiving end, again simplistically, that says that when the receiving system gets coordinates with R=G=B for the color of a pixel in the image, the display should generate light whose chromaticity is that of the white point of the color space.

But at the encoding end (i.e., the camera), although you don't see it described this way (except by moi), our real intent is not that the coordinates describe the color of the light from each point in the image, but rather that they describe the reflective color of that spot in the scene. So for our "white" note paper, we want the encoding (if in sRGB) to be R=G=B, in all situations of illumination.

Then, if in fact the illumination on the area surrounding the display unit in the viewing room has the same chromaticity as the white point of the color space, the eye will always see that piece of "white" paper as "white", regardless of the illumination under which it was photographed.

How do we make that happen? The camera is not in a position to ascertain the reflective chromaticity of points in the image - it can only record the color of the light reflected from them. Well the "classical" way is to include in the scene (for the actual shot, or a test shot) an item of "neutral" reflectance (such as a "gray card", or the white pendant I have Carla wear for the purpose.

Then, when we are developing the raw camera output in our raw converter, we point to that item with the eyedropper, and tell the program, "This spot is white [in reflective color], by golly, and I want to you encode it as white" - that is, if the output color space is sRGB, to encode it with coordinates for which R=G=B. (And the corresponding thing will then follow for other chromaticities in the captured image.)

[To be continued]
 

Doug Kerr

Well-known member
[Part 3]

Now suppose when developing a new RGB-family color space we decide that it would be practical to base it on the "intended" chromaticity of the light on the surrounding areas of the viewing room corresponding to standard illuminant D50. Then, we would adopt that same chromaticity for the white point of the color space, and work out the rest of the details from that. And this would make The Story work out just fine.

But there was a clinker, I haven't mentioned it so far, but one of the overall objects of a complete color management system would be to get visual consistency between paper prints and images seen on a display. (Perhaps adjustments are made in color as seen on a display, with the object that the printed image would look like that.)

Now when we consider viewing a print, we can really be very casual about the matter of the chromaticity of the light in the viewing room. The situation is just like that of direct observation of our white piece of paper under different illuminants. The chromaticity of the illumination indeed affects the chromaticity of the light reflected by any spot on the print, but also it also affects the chromaticity of the light reflected from familiar surrounding objects that the human visual system uses to reckon the reflective color of the items in which we are interested. (It's actually not quite that tidy, but that will do for our purposes here.)

Now we can (subject to some imitations) cause a tricolor display to exhibit any chromaticity we want for "white". But it doesn't work that way for a (simple) printer. For "white", the printer (simplistically) does not deposit any ink at all. Thus the reflective color of any "white" spot in the printed image is the reflective color of the "bare" paper stock.

Now at the time the sRGB color space was being developed, it turned out that commonly-available good glossy printer papers typically were a little on the "blue" side of spectrally uniform. (The reason? To make prints look "brighter", the same reason that laundry detergents have blue UV-sensitive fluorescent dyes in them.) The result of this in the context of The Story is that an ideally-rendered displayed image would look a little "yellowish" compared to an ideally-rendered print.

The solution? Just what any engineer in a rush to go to lunch would do - plug it. So even though it had been decided that the preferred chromaticity for the illumination on the surrounding areas in a (display) viewing room was "D50", the white point of the color space was made "D65" (a slightly more "bluish" chromatically). And thus, when everything was working as intended, the print and display image would look very much the same.

-#-​

Best regards,

Doug
 

Tom dinning

Registrant*
Another masterpiece from the engineer. (by the way, Doug, we call train drivers engineers here, just as a maintenance man in the UK might be called an engineer)

Trouble is, I've had 2 glasses of fine wine and none of it made any sense except the last two lines, which is exactly what I do but don't know why.

When I'm sober I'll read the rest. Or perhaps not, since all that knowledge will make no difference to my decision making.

xx
tommy
 

Doug Kerr

Well-known member
Hi, Tom,

Another masterpiece from the engineer. (by the way, Doug, we call train drivers engineers here, just as a maintenance man in the UK might be called an engineer)

Yes, as is the case here. But I needed to make the distinction when I wrote to you about my style of dress.

Trouble is, I've had 2 glasses of fine wine and none of it made any sense except the last two lines, which is exactly what I do but don't know why.

Well, I'm sorry to have not included in a preface recommendations as to what king of wine and how much to drink before reading the article.

When I'm sober I'll read the rest. Or perhaps not, since all that knowledge will make no difference to my decision making.

Makes sense to be.

Best regards,

Doug
 
Hello Doug, an interesting article, thank you.

Could you please address the case where the color being imaged is self-excited and surrounded by darkness? For example, a lone traffic light out in the countryside in overcast conditions at dead of night, to be displayed on a perfect sRGB monitor.

Thanks,

Ted
 

Doug Kerr

Well-known member
Hi, Ted,

Hello Doug, an interesting article, thank you.

Could you please address the case where the color being imaged is self-excited and surrounded by darkness? For example, a lone traffic light out in the countryside in overcast conditions at dead of night, to be displayed on a perfect sRGB monitor.

You raise an important point. I have pondered this from time to time, never reaching a fully-comfortable conclusion!

I wouild seem that in this case, the human visual system having no basis (in terms of "surrounding objects") to try and conclude the "reflective color" of the object (which of course it does not have in this case, it being self-luminous), the visual system probably operates on the basis of some "default" presumption about the ambient light.

It would seem that in this case we would do the best by attempting to encode the traffic light so its chromaticity in the color space we use matches its "actual" chromaticity.

But how would we tell our camera to do that (or tell our raw converter to do that)? It is tempting to say, "Well, just set 'no' white balance color correction." But of course there is no such thing (and certainly no such setting on the controls of our camera or of our raw converter). The camera always bases its image processing on 'some" presumed chromaticity of the ambient illumination.

After contemplating this conundrum a few years ago, I concluded that what we need to do is to set the "white balance color correction" to a preset for incident light whose chromaticity is that of the white point of our color space.

So if that is sRGB, whose white point is the chromaticity of standard illuminant D65, and our camera offers a white balance color correction in terms of "Kelvin", can we set that to 6500 and be there? Well maybe. We recall that a chromaticity is not fully specified by a correlated color temperature (CCT), as there can be many chromaticities that exhibit a certain CCT.

So when we set out camera to a white balance color correction of "6500K", does that means that (if we do not tamper with the "offset", if we even can) the presumption is of the "black body" chromaticity with that CCT (which is not quite that of illuminant D). Well, we may not know, exactly.

Now I am of course "splitting hairs" here, as I think it is best to understand the subtleties even if we don't have to be fully guided by them in reasonable practice. That having been said, I suspect that if on our camera we set the white balance color correction to "6500K" we will be pretty close (for the sRGB color space) to having 'no' color correction (this is after all, not a precision colorimetric measurement we are trying to make - just to get a picture of a traffic signal aspect).

Best regards,

Doug
 
Top