AWB Test

[Colin Southern]

Hi Paul,

Manfred tried to explain to me as well that RAW is not affected by WB settings. No matter what WB setting you use in RAW you can change it in PP. A very good and valid reason to shoot in RAW.
I have to agree with you as far as getting it right in camera. I believe there is nothing more satisfying than getting it right in camera.
I shoot a lot, using manual WB, the great danger is forgetting to set WB back to Auto. If you are shooting in JPEG the shots will be ruined.
I must admit, RAW is a bit like a life jacket on a rubber duck on the open see. If you don't use it and fall off, the chance of drowning is too big to risk. Unless of cause you are a Michael Phelps.

We will never get to old to learn.

Hi Andre,

You can really only "get it right in camera" if you're shooting JPEG - a RAW capture can't be white balanced until it's converted & "reassembled" - so all a white balance setting on the camera is doing is white balancing the in-camera JPEG that you see on the review screen and giving the RAW converter a suggested place to start.

[ajohnw]

It's a curious thing that cameras awb often get incandescent lighting wrong and usually warm things up. For tungsten I am not at all surprised that the white card produced a figure of 2,600K. That is the sort of figure that would be expected from tungsten lamps. The usual figures quoted are around that maybe going as high as 2,800K. Halogen is higher. I usually use awb and camera white balance if developing raw but often artificial lighting causes problems but not on all shots. It must be down to the colour mix / light level variations. I shudder to think what some LED lighting would do. There is so much excess blue coming out of near daylight types it's bound to cause problems - it does on a microscope but colour balance is often way out when cameras are used on these unless there is some semblance to normal photo's.

-

[oleleclos]

I agree about RAW not being colour corrected until at the conversion stage, but I believe the reason Paul's AWB setting renders this shot warmer than the other two could be intentional.

My Nikon (D800) has two options on the AWB menu: "Normal" and "Keep warm lighting colors". I haven't checked but I would expect the RAW converter to do more or less what the in-camera JPEG converter does with these settings.

The former should try to fully correct for the colour temperature of the light source but the latter will keep at least some of the warmth from indoor lighting to match the way many perceive such a scene. Not sure what AWB options Paul's camera has.

Graham's sequence of pictures beautifully illustrates that not everything should be fully "corrected".

[AB26]

Hi Andre,

You can really only "get it right in camera" if you're shooting JPEG - a RAW capture can't be white balanced until it's converted & "reassembled" - so all a white balance setting on the camera is doing is white balancing the in-camera JPEG that you see on the review screen and giving the RAW converter a suggested place to start.

Technically that might be correct. However if I shoot in RAW and do WB adjustments and/or adjustments to tone and/or contrast, what I see when opening the “RAW” files in View NX2 is different with different settings. All “ RAW” images from my Nikon do not look like they were shot in Auto mode.
In View NX2 I am able to change WB, contrast, boost colour etc. or I can leave it as it was transferred from the camera, before “converting” the file to JPEG or TIFF. Technically that means I can adjust images by adjusting camera settings, shot in RAW or JPEG, to render the image as close to what I want it to look like, SOOC.

A "RAW" image cannot be opened by Picasa. I would have liked to see if there is any change if opened in different programs.

[Melkus]

I have come to the conclusion at less for me to shot in RAW like I do and leave the WB for most part on Auto then during pp to adjust it as needed.

[Colin Southern]

Technically that might be correct. However if I shoot in RAW and do WB adjustments and/or adjustments to tone and/or contrast, what I see when opening the “RAW” files in View NX2 is different with different settings.

For sure. Manufacturers conversion products generally honour all picture style tags, whereas the likes of ACR only carry white balance through - in all cases they're just starting points though; the actual data isn't affected, so in essence all that the tags do is "pre-adjust" the adjustment sliders for you.

[John Morton]

I'd like to chime in on a couple of points.

First, the idea of "getting it right in-camera": now, although we can indeed say that the RAW file data is always 'correct' and 'is what it is' the idea of getting the white balance correct when the photo is being shot is still valid.

Consider: we are talking about setting the color information of a photograph correctly, so, we must ask ourselves - correct with reference to what? well, the scene and lighting conditions being imaged, obviously! So... how exactly do we insure that the settings for white balance which we come up with after the fact are accurate AFTER we have left the very scene upon which we would necessarily have to base our judgments?

Any white balance settings arbitrarily determined for a RAW file AFTER the scene being photographed has been left behind are necessarily accurate ONLY IN THEORY. The ONLY way to determine an actual and accurate white balance setting for a RAW file after the scene photographed has been left is through a know quantity (white balance card or color chart) included in a photograph taken under the original conditions.

In philosophy, this distinction is one which is made between the 'soundness' (logical consistency) of an argument and its 'validity (direct relation to actually occurring circumstances). That the white balance of a RAW file can be set at any point after it has been obtained is a SOUND argument; but, that does not establish the VALIDITY of such a determination with respect to the original situation in which the photograph was taken.

So; out the other end of that argument, we have a basic observation which I think all photographers can relate to at one point or another: it is a lot easier to take an accurate in-camera white balance and to then tweak a RAW file until the image meets out expectations, than it is to take an inaccurate in-camera white balance and then try to bring the colors back to what they should have been in the first place...

[xpatUSA]

Are there not, in many scenes, objects for which the color (in terms of hue, saturation or a*,b* or u*,v* or CIE x,y) is known? If so, a card may not be strictly necessary. I refer to objects such as white houses, billboards, ordinary concrete, skies or clouds (taking into account time of day, latitude, etc), or even lamp-shades ;-).

So, if one makes make a blue sky the correct blue (by global correction to the color space) - do all the other hues in the scene magically come into line, I wonder?

A good test would be to shoot a MacBeth card with the wrong WB set in-camera. Since the card colors are published and known, does correcting one of the color patches automatically reward one with a row grayscale patches along the bottom?

Failing that, maybe one's latest editor version can outperform the in-camera AWB function. The lamp scene in the OP, with it's lack of saturated color and low contrast would probably fool a simple AWB quite easily, I reckon :-)

[Colin Southern]

Are there not, in many scenes, objects for which the color (in terms of hue, saturation or a*,b* or u*,v* or CIE x,y) is known? If so, a card may not be strictly necessary. I refer to objects such as white houses, billboards, ordinary concrete, skies or clouds (taking into account time of day, latitude, etc), or even lamp-shades ;-).

Unfortunately, the degree to which various object reflect differing wavelengths isn't a constant. It's called Illuminant metameric failure.

[Colin Southern]

Isn't that to do with an object's differing color appearance under different illuminants? Whereas, the illuminant in the OP is the same in all the shots.

Yes - and as a result its impossible to predict what correction is going to be required without knowing either the colour temperature or the spectral response characteristics of the object - and unfortunately, you're unlikely to know either - thus you have 2 variables, not just one.

The beauty of cards like the WhiBal is not just that they're spectrally neutral - they also don't suffer from significant metameric issues.

If you had a known object that was known not to have metameric variations then your idea would work, but even then I doubt many would either have details of colour info for an object or a photospectrocolorimeter handy to measure it.

[Glenn NK]

Hi Paul,I have to agree with you as far as getting it right in camera. I believe there is nothing more satisfying than getting it right in camera.

In film days it wasn't only satisfying, it was important.

With digital RAW, getting it right in the camera isn't nearly as important, and some would argue that it's not important at all.

Personally I never change the AWB settings on my camera - shooting RAW exclusively, it isn't worth the bother.

Glenn

[Letrow]

Interesting experiment Paul. Everything has been said already, but it is still fun to look at the photos and see what you can do with them.
I sometimes find that a white card is fine, but sometimes gives too cool of an end result, depending on the lighting that you have in certain rooms. In that case it is nice that you can easily make adjustment and bring it up to your own taste.

[xpatUSA]

Yes - and as a result its impossible to predict what correction is going to be required without knowing either the colour temperature or the spectral response characteristics of the object - and unfortunately, you're unlikely to know either - thus you have 2 variables, not just one.

The beauty of cards like the WhiBal is not just that they're spectrally neutral - they also don't suffer from significant metameric issues.

If you had a known object that was known not to have metameric variations then your idea would work, but even then I doubt many would either have details of colour info for an object or a photospectrocolorimeter handy to measure it.

There could be a misunderstanding. The OP shows three shots of a scene with but one source of light. I don't understand how metamerism, or the earlier mention of "illuminant metameric failure", is relevant to that case. [edit: I suppose it is relevant to using the same reference object indoors or outdoors, so to speak]

But, as far as my earlier post is concerned, you're right - only a hopeless photogeek would open the image in L*a*b* and try a delta-E correction!

[Colin Southern]

There could be a misunderstanding. The OP shows three shots of a scene with but one source of light. I don't understand how metamerism, or the earlier mention of "illuminant metameric failure", is relevant to that case. [edit: I suppose it is relevant to using the same reference object indoors or outdoors, so to speak]

If you're still referring to white balancing by applying a set correction to an object who's colour information is known (which is what my reply was addressing) then illuminant metameric failure could mean, and probably would mean (depending on how accurate you needed to be) that the correction required would be slightly different for different items that you had information for.

Eg my RRS bracket appears black under whiteish/neutral light (so in LAB it would probably be close to something like 10,0,0), but under a tungsten illuminant it appears purple (I wouldn't even like to guess at the LAB values) - and yet the camera it's attached to appears black under both illuminants - and would probably have similar LAB values (at least in the AB channels). If one tried to correct an image based on colour values of the bracket, I suspect that the results would vary widely depending on the illuminant used at the time of capture, compared to the values for the camera body itself.

[xpatUSA]

If you're still referring to white balancing by applying a set correction to an object who's colour information is known (which is what my reply was addressing) then illuminant metameric failure could mean, and probably would mean (depending on how accurate you needed to be) that the correction required would be slightly different for different items that you had information for.

Eg my RRS bracket appears black under whiteish/neutral light (so in LAB it would probably be close to something like 10,0,0), but under a tungsten illuminant it appears purple (I wouldn't even like to guess at the LAB values) - and yet the camera it's attached to appears black under both illuminants - and would probably have similar LAB values (at least in the AB channels). If one tried to correct an image based on colour values of the bracket, I suspect that the results would vary widely depending on the illuminant used at the time of capture, compared to the values for the camera body itself.

The bracket sounds quite interesting as an example. What material is it made of and what kind of finish? Is the purple color constant irrespective of the angle of the lighting unlike, for example, an oil film?

A specular specular reflection on the base of the statuette looked a likely candidate - one click in PSE6's "Remove color cast" produced the following:



Much too blue but other points on the base did better than that - the perfect one showed a nice whitish yellow on the lamp shade until my shaky finger double-clicked it out of existent! Oddly, the white rail on the wall was much less successful - metameric paint, perhaps? (just kidding . . .)

[Colin Southern]

The bracket sounds quite interesting as an example. What material is it made of and what kind of finish? Is the purple color constant irrespective of the angle of the lighting unlike, for example, an oil film?

It's an alloy of sorts - probably best described as "aircraft grade", but other than that I wouldn't know. It's typical RRS quality - eg "top shelf" - not "muck metal" like so many other cheap components are made out of. The finish is matt and "anodised"? Dunno - it's been some 30+ years since I studied metalurgy. Yes - constant with respect to angle of the lighting.

A specular specular reflection on the base of the statuette looked a likely candidate - one click in PSE6's "Remove color cast" produced the following:



Much too blue but other points on the base did better than that - the perfect one showed a nice whitish yellow on the lamp shade until my shaky finger double-clicked it out of existent! Oddly, the white rail on the wall was much less successful - metameric paint, perhaps? (just kidding . . .)

Sorry, but I'm really not sure what you're trying to achieve here. Yes - one can get what - visually - appears to be a "good" result using a specular reflection (so long as it doesn't have a blown channel) - but it really depends on the colour of the object and the colour of the light, so it's nothing scientific and really amounts to nothing more than just a candidate to try and visually correct mystery "lighting", with other candidates being probably the wall and perhaps the base of the lamp. With a KNOWN spectrally neutral object (such as a WhiBal card) then one can simply white balance a shot without any kind of particular evaluation being necessary. In my case - for studio portraiture anyway - I usually shoot a colour passport (both grey card and 24 patch) - produce a profile - apply the profile - white balance to the grey card - add a couple of hundred kelvin (to warm skin tones) and "job done". Quick and easy and takes less than a minute.

[xpatUSA]

Sorry, but I'm really not sure what you're trying to achieve here.

In my muddled way I was suggesting that, in the absence of KNOWN spectrally neutral objects, a neutral-ish object in a scene could be used in a pinch. It seemed to me that there were two such objects in the OP scene.

Always nice of course to have a WhiBal card handy for any occasion, but I'm not a pro and don't even carry my mini-Macbeth card with me at all times (blush).

Yes - one can get what - visually - appears to be a "good" result using a specular reflection (so long as it doesn't have a blown channel) - but it really depends on the colour of the object and the colour of the light, so it's nothing scientific and really amounts to nothing more than just a candidate to try and visually correct mystery "lighting", with other candidates being probably the wall and perhaps the base of the lamp. With a KNOWN spectrally neutral object (such as a WhiBal card) then one can simply white balance a shot without any kind of particular evaluation being necessary. In my case - for studio portraiture anyway - I usually shoot a colour passport (both grey card and 24 patch) - produce a profile - apply the profile - white balance to the grey card - add a couple of hundred kelvin (to warm skin tones) and "job done". Quick and easy and takes less than a minute.

Quite a fast worker, then ;-). Seriously, after some study of metamerism prompted by your good posts, it appears that my original supposition might hold but only for metameric objects, if I read you right. Thus if we were to shoot, say, a Macbeth card, a Kodak graycard, a WhiBal card all in one scene but under two or three different illuminants with appropriate WB selections they should all look the same or, with incorrect WB, at least have the same color cast. (assuming the cards are of equal metameric performance).

Then, if some other object in a scene performed equally (by prior experiment or experience) - it would be a worthy candidate, of course much easier if it were a neutral color: concrete, white paint, dead wood, ashes, and such.

Thank you for pointing out the perils of Illuminant Metameric Failure!

[Colin Southern]

In my muddled way I was suggesting that, in the absence of KNOWN spectrally neutral objects, a neutral-ish object in a scene could be used in a pinch. It seemed to me that there were two such objects in the OP scene.

Yes, of course - that's exactly what we do - just not sure how that ties in with:

Are there not, in many scenes, objects for which the color (in terms of hue, saturation or a*,b* or u*,v* or CIE x,y) is known? If so, a card may not be strictly necessary. I refer to objects such as white houses, billboards, ordinary concrete, skies or clouds (taking into account time of day, latitude, etc), or even lamp-shades ;-).

So, if one makes make a blue sky the correct blue (by global correction to the color space) - do all the other hues in the scene magically come into line, I wonder?

which is what I've been addressing.

Quite a fast worker, then ;-).

Not particularly - one only needs to drop the reference shot into the ColourPassport - name the resultant profile - and then batch apply the new profile to all shots in the series - so only a few mouse clicks.

Seriously, after some study of metamerism prompted by your good posts, it appears that my original supposition might hold but only for metameric objects, if I read you right. Thus if we were to shoot, say, a Macbeth card, a Kodak graycard, a WhiBal card all in one scene but under two or three different illuminants with appropriate WB selections they should all look the same or, with incorrect WB, at least have the same color cast. (assuming the cards are of equal metameric performance).

I would expect them to be close - but then again - that's pretty much what they were designed to do.

Then, if some other object in a scene performed equally (by prior experiment or experience) - it would be a worthy candidate, of course much easier if it were a neutral color: concrete, white paint, dead wood, ashes, and such.

Yes - in theory. In practice though I suspect that most would find others ways are easier to produce an acceptable result with.