White balance and exposure determination

[Doug Kerr]

Drew Strickland, in a new paper entitled "Busting the Top 7 White Balance, Color Balance and Grey Balance Myths", introduces an interesting concept regarding the interaction of white balance color correction technique and exposure determination.

In one section of his paper, "Myth #2: Raw is the Answer ", he gives a rationale as to why making white balance color correction during external development of the raw data is inferior to using in-camera white balance color correction. As I understand it, the rationale goes like this:

• Suppose we plan to take the raw output from the camera and do white balance color correction during external raw development.
• We have the camera's white balance control set to some arbitrary preset, or perhaps to AWB.
• We plan to set exposure based on observation of the camera tricolor histogram on a test shot, basically using the "expose the hottest channel to the right" criterion. We note that this histogram is influenced by the application, in-camera, of the "arbitrary" white balance color correction that is set.

We then load the file for the actual shot into our raw development software, and enact the desired white balance color correction. We may find that the highest of the three histogram peaks is now not at the far right. (Or is "beyond" the far right - Drew's example, however, is for the former situation.)

Thus we find that we have, in effect, "underexposed" the actual shot. While we can compensate for this by elevating the image levels in the raw development software ("pushing" the development), doing so will engender a noise penalty.

Thus, says Drew, if we were set the camera's custom white balance setting based on a measurement (with a white balance measurement diffuser or such), then the camera histogram would be a better indicator of the "headroom" (on a three-channel basis) for the proposed shot as corrected, leading to an exposure that will result in better noise performance for the color-corrected image.

I would be interested in the outlooks of those here who are more familiar with these techniques than I on the impact of the camera white balance setting on the attainment of "proper" exposure by chimping the histogram (for either a single or three-channel histogram).

As a plus, those who are interested in what "grey balance" means will find, in the paper, that it is a synonym for "gray balance".

Drew's paper can be found here:

http://www.prophotohome.com/busting-...-balance-myths

Here is my discussion with Drew over the concept I mention:

http://www.prophotohome.com/forum/wh...tml#post470737

[Bill Miller]

Doug,

Why waste your time on Drew - He is only trying to sell is own product. Look at it this way he took a great set of forums and ruined it. Does he have any creditability other then what he gives himself?

Save your breath!

[Doug Kerr]

Hi, Bill,

Originally Posted by Bill Miller

Doug,

Why waste your time on Drew - He is only trying to sell is own product. Look at it this way he took a great set of forums and ruined it.

Indeed. But I am nevertheless intersted in the question I posed.

Thanks.

Best regards,

Doug

[Andrew Rodney]

Originally Posted by Bill Miller

Doug,
Why waste your time on Drew - He is only trying to sell is own product. Look at it this way he took a great set of forums and ruined it. Does he have any creditability other then what he gives himself?

Amen. I put my two cents in but I agree, this guy is just a zealot.

[Alain Briot]

Doug,

If there's really a risk of under or over exposing because of an inaccurate white balance setting, simply bracketing the exposure will take care of this issue. Bracketed exposures are free, highly recommended, and always good to have.

Why make things more complicated than they have to be? Simplicity is a virtue!

Furthermore, in fine art landscape photography, there's no such thing as an "exact" white balance. White balance is whatever I like the image to look like. So I am bound to change it from exposure to finished image, and therefore bracketing is the answer to all issues, white balance and other.

[Andrew Rodney]

I don't see how exposure (or lets say proper let alone idealized exposure) has anything to do with white balance for Raw data.

[Alain Briot]

Andrew,

I don't know either, but if it does then bracketing would take care of it. I bracket anyways so this takes care of all problems related to exposure, whatever they may be.

[Andrew Rodney]

I'm all for bracketing unless I'm shooting portraits.

[Bart_van_der_Wolf]

Originally Posted by Doug_Kerr

I would be interested in the outlooks of those here who are more familiar with these techniques than I on the impact of the camera white balance setting on the attainment of "proper" exposure by chimping the histogram (for either a single or three-channel histogram).

Originally Posted by Alain Briot

Doug,

If there's really a risk of under or over exposing because of an inaccurate white balance setting, simply bracketing the exposure will take care of this issue. Bracketed exposures are free, highly recommended, and always good to have.

While I agree about the bracketing in general (if one wants ultimate quality one will need to bracket a lot), Doug's technical understanding raised a valid wondering, and sparked by his technical understanding of fundamental physics, it's more to the point than many would grasp.

Originally Posted by Andrew Rodney

I don't see how exposure (or lets say proper let alone idealized exposure) has anything to do with white balance for Raw data.

Ah, not so fast! Exposure quality itself will benefit from proper exposure across the spectral frequencies in the scene at hand. This means that it will pay off to color/colour balance to the dominant lightsource (!), or more correctly "illuminant". The best way to achieve this is by using proper filters on the lens, to match the scene's color balance to the sensor's sensitivity. This will allow to expose the R/G/B channels to "the right" and achieve a high Signal to Noise (S/N) ratio. This will allow an optimal signal quality to 'play' with in post-processing.

Just to be clear, the in-camera color balance setting is basically a post-processing clue when shooting Raw, but adjusting the spectral quality of the scene's actual photometric exposure will allow to improve the image's technical quality. It will influence the in-camera's JPEGs, not its Raw data.

Bart

[Doug Kerr]

Hi, Andrew,

Originally Posted by Andrew Rodney

I don't see how exposure (or lets say proper let alone idealized exposure) has anything to do with white balance for Raw data.

I don't know how either. I certainly never said anything about the effect of exposure on white balance color correction.

The issue I am discussing is whether differences in the camera's white balance setting can influence the exposure decision the photographer might make when choosing exposure by observation of the camera histogram on a test shot.

Best regards,

Doug

[Daniel Buck]

it may be different on other cameras, but on my 1Ds2 the white balance for most situations is usually pretty close (pretty close to what I want it to be anyway). I think the histogram will only read way off if the white balance is completely wrong (off by thousands of kelvin). if it's close like the auto why balance usually seems to be in most situations, then I think you're ok for looking at the histogram.

An interesting point to bring up though, but I doubt it would seriously affect things in most situations if the AWB is relativly close.

[Andrew Rodney]

Originally Posted by Bart_van_der_Wolf

Ah, not so fast! Exposure quality itself will benefit from proper exposure across the spectral frequencies in the scene at hand. This means that it will pay off to color/colour balance to the dominant lightsource (!), or more correctly "illuminant". The best way to achieve this is by using proper filters on the lens, to match the scene's color balance to the sensor's sensitivity.

But that's not what we're discussing here or on that forum (and I agree, using a filter could help). We're talking about creating a metadata tag from a product to tell the converter something so it can apply a different rendering default. IOW, I'm asking how one metadata WB tag versus another has any effect on the actual Raw data. My point was (excluding putting something between the lens and the sensor), exposure and ISO affect the actual Raw data, not the WB tag that this product creates for users.

[Doug Kerr]

Hi, Andrew,

Originally Posted by Andrew Rodney

We're talking about creating a metadata tag from a product to tell the converter something so it can apply a different rendering default.

I have to again emphasize that my discussion here has nothing to do with the camera WB setting metadata tag, or with doing rendering during raw development based on it.

Certainly making different WB settings in the camera makes no change on the delivered raw data if a consistent camera exposure is used.

The matter I discussed is how different WB settings on the camera can cause the photographer, for a given scene, to choose different camera exposures if he makes that choice based on examination of the camera histogram on test shots. And of course the raw data is different for shots of the same scene taken with different camera exposure.

I understand your view that making exposure decisions that way is not good practice. From our dialog in another forum, I think your preference is to shoot at a thoughtfully metered exposure (and/or use bracketing, where practical).

But evidently many photographers use that practice, and I was intersted in a matter that affects it.

Best regards,

Doug

[Mike Shimwell]

Originally Posted by Doug_Kerr

Hi, Andrew,


I don't know how either. I certainly never said anything about the effect of exposure on white balance color correction.

The issue I am discussing is whether differences in the camera's white balance setting can influence the exposure decision the photographer might make when choosing exposure by observation of the camera histogram on a test shot.

Best regards,

Doug

Doug

The easiest response to your question reflects Drew's comments, but the asnwer does not need his piece of kit. The in camera white balance setting can mislead you when you try to expose to the right. The (in camera) white balance setting adjusts the multiplier for each channel and so may lead to one channel blowing out even though the raw data was not clipped. By adjusting exposure to allow for this you will underexpose compared to a full expose to the right exposure.

See below for a possible approach to address this issue.

Mike

[Doug Kerr]

For the benefit of those who joined the discussion late, let me summarize the matter I originally introduced here. I am using more complicated language than I did at first, since various dialog has given me some greater insight into the matter than I first had.

It has been suggested by Drew Strickland that, rather than:

• Leaving the camera WB control set to some random choice, shooting, taking the raw output, and applying white balance color correction during raw development (based on any of various premises).

it is better to:

• Develop a Custom White Balance setting in the camera (based on, for example, the use of a white balance measurement diffuser or a calibration shot on a neutral target), shoot with that CWB in effect, and then either:
•• Take the (white balance color corrected) JPEG output,
or
•• Take the raw output and apply white balance color correction during raw development (possibly based on the camera CWB setting or possibly not).

The rationale for this recommendation is this (my comments in red):

• The camera histogram is based on an RGB image the camera generates from its raw data, applying color correction in accordance with the current camera WB setting. Evidently true.
• For a given scene with a given camera exposure, the histogram will vary with different WB settings. True (I tested it).
• If the photographer makes his exposure decision for the actual shot based on examination of the histogram for test shots, he may make a different exposure decision for one WB setting than another. I can believe that.
• The exposure decision made based on examination of the histogram with the "appropriate" WB setting in effect (CWB, based on scene measurement) will probably be a "more appropriate" exposure than that made based on the examination of the histogram with some other WB setting in effect. In fact, the exposure chosen based on the "arbitrary" WB setting will usually be less than the one with the applicable CWB WB setting, so this difference constitutes an "underexposure". Here's where I get off the train, guys.

Best regards,

Doug

[Doug Kerr]

Hi, Mike,

Thanks for your insights, which make sense to me.

Originally Posted by Mike Shimwell

See below for a possible approach to address this issue.

The link as it arrived here just takes me to your home page. How can I locate your piece on this topic?

Thanks.

Best regards,

Doug

[Mike Shimwell]

Originally Posted by Bart_van_der_Wolf

While I agree about the bracketing in general (if one wants ultimate quality one will need to bracket a lot), Doug's technical understanding raised a valid wondering, and sparked by his technical understanding of fundamental physics, it's more to the point than many would grasp.



Ah, not so fast! Exposure quality itself will benefit from proper exposure across the spectral frequencies in the scene at hand. This means that it will pay off to color/colour balance to the dominant lightsource (!), or more correctly "illuminant". The best way to achieve this is by using proper filters on the lens, to match the scene's color balance to the sensor's sensitivity. This will allow to expose the R/G/B channels to "the right" and achieve a high Signal to Noise (S/N) ratio. This will allow an optimal signal quality to 'play' with in post-processing.

Just to be clear, the in-camera color balance setting is basically a post-processing clue when shooting Raw, but adjusting the spectral quality of the scene's actual photometric exposure will allow to improve the image's technical quality. It will influence the in-camera's JPEGs, not its Raw data.

Bart

There are a couple of approachses that can be used to optimise exposure - likely together.

You can determine the white balance setting that applies the same multiplier to each raw channel and force the in camera white balance to this setting. By then setting shot parameters to minimse contrast and saturation boost you will get jpg histograms as close as possible to the raw histogram (so to speak, please don't call me on this slightly loose description) and can optimise expsoure to avoid clipping any channel.

Going on from here, as Bart suggested, by filtering in front of the lens to equalise the response of each channel (typically requiring a magenta filter) it is possible to equalise each raw channel and then to undo in processing. This can minimse noise in the final image. Some studio photographers filter their flashes with magenta filters to achieve the same end.

Practically, I tend to expose to the right relying on the green channel as provided that isn't blown it is unusual for the red or blue raw channels to blow in normal light - I till bracket though if it's reasonably easy to do.

Mike

[Asher Kelman]

Doug,

Could you expand the point where you "get of the train". How do you define "CWB".

Originally Posted by Doug_Kerr

• The exposure decision made based on examination of the histogram with the "appropriate" WB setting in effect (CWB, based on scene measurement) will probably be a "more appropriate" exposure than that made based on the examination of the histogram with some other WB setting in effect. In fact, the exposure chosen based on the "arbitrary" WB setting will usually be less than the one with the applicable CWB WB setting, so this difference constitutes an "underexposure". Here's where I get off the train, guys.

Best regards,

Doug

I would like to point out that the histogram shown on the camera LCD, based on whatever setting, may still have more room to the right even when it abuts the right. We need to check that with the new cameras.

Asher

[Daniel Buck]

Originally Posted by Doug_Kerr

It has been suggested by Drew Strickland that, rather than:

• Leaving the camera WB control set to some random choice, shooting, taking the raw output, and applying white balance color correction during raw development (based on any of various premises).

it is better to:
...
•• Take the raw output and apply white balance color correction during raw development (possibly based on the camera CWB setting or possibly not).

What's the difference in those two? Sounds the same to me, setting the white balance during raw converting.

[Asher Kelman]

Originally Posted by Bart_van_der_Wolf

While I agree about the bracketing in general (if one wants ultimate quality one will need to bracket a lot), Doug's technical understanding raised a valid wondering, and sparked by his technical understanding of fundamental physics, it's more to the point than many would grasp.

Bart,

You are right on here! Doug should be commended. We should not take the design guesses or compromises made by the camera MFRs/WB salesmen as being in the best interests of professional photographers. We need to know the what, where, how and wherefore of these compromises. Thanks Doug.

Originally Posted by Bart_van_der_Wolf

Ah, not so fast! Exposure quality itself will benefit from proper exposure across the spectral frequencies in the scene at hand. This means that it will pay off to color/colour balance to the dominant lightsource (!), or more correctly "illuminant". The best way to achieve this is by using proper filters on the lens, to match the scene's color balance to the sensor's sensitivity. This will allow to expose the R/G/B channels to "the right" and achieve a high Signal to Noise (S/N) ratio. This will allow an optimal signal quality to 'play' with in post-processing.

This is a brave approach and could be done in camera with a set of filters just like the built in ND filter in the Canon G10 digicam. I'd love to hear more on this. This seems better than anything else I've heard, including bracketing.

Originally Posted by Bart_van_der_Wolf

Just to be clear, the in-camera color balance setting is basically a post-processing clue when shooting Raw, but adjusting the spectral quality of the scene's actual photometric exposure will allow to improve the image's technical quality. It will influence the in-camera's JPEGs, not its Raw data.

Bart,

Explain this last point.

"but adjusting the spectral quality of the scene's actual photometric exposure". If it's, indeed, the actual photometric exposure, which has been optimized by either

• the lighting or
  
  
• the use of correction filters

why then would it only apply to the JPG made by the camera and not the RAW file data?

Is it that by your designation, "but adjusting the spectral quality of the scene's actual photometric exposure" you mean merely the "math" corrections based on the chosen in-camera "appropriate" scene setting such as Tungsten, Daylight etc?

Asher

[Andrew Rodney]

Originally Posted by Doug_Kerr

• The camera histogram is based on an RGB image the camera generates from its raw data, applying color correction in accordance with the current camera WB setting. Evidently true.

For the JPEG based on in-camera processing. I agree.

• For a given scene with a given camera exposure, the histogram will vary with different WB settings. True (I tested it).

Again, for the JPEG, again, I agree.

• If the photographer makes his exposure decision for the actual shot based on examination of the histogram for test shots, he may make a different exposure decision for one WB setting than another. I can believe that.

For the JPEG, fine. My beef with Drew is he's now suggesting that a photographer use this big fat lie on the camera LCD as an indicator for exposing the Raw data. That's simply a dumb thing to do as the Histogram, with or without Drew's product has no bearing on the Raw data. Its like suggesting that this histogram will match what we get in Photoshop after we edit the JEPG. Probably will not. Or suggesting the color on the LCD has a bearing on what the final output on another emissive display will look like. Most likely no way.

• The exposure decision made based on examination of the histogram with the "appropriate" WB setting in effect (CWB, based on scene measurement) will probably be a "more appropriate" exposure than that made based on the examination of the histogram with some other WB setting in effect. In fact, the exposure chosen based on the "arbitrary" WB setting will usually be less than the one with the applicable CWB WB setting, so this difference constitutes an "underexposure". Here's where I get off the train, guys.

And I submit, if you use this information as a guide to the Raw data, you're fooling yourself. "More appropriate" could be said to be "less a lie" or "less incorrect". But this is the improper use of the tool, that's the bottom line. Its the improper understanding of what the Histogram is showing you based on the data you will end up with. Its a delusional, make yourself feel better button. If you're shooting Raw, if your goal is to ETTR or have some actual idea of the effect of exposure on Raw data, looking at an output referred, gamma corrected histogram, based on in camera JPEG processing is far from accurate, useful and a good use of that visual feedback. Producing a metadata tag that alters that histogram and has no effect on the Raw isn't useful. And worse, to write a piece supposedly exposing myths to sell a product, while in effect producing a myth in the process is inexcusable.

All Drew needs to do is provide a piece that actually explains when and why his product is beneficial; for the creation of JPEGs in camera. He should leave Raw out of the discussion. Trying to convince users that making a new lie to make the original lie a bit less a lie is far from truthful.

[Doug Kerr]

Hi, Asher,

Originally Posted by Asher Kelman

Could you expand the point where you "get of the train". How do you define "CWB".

Sorry - "custom white balance".

I get off the train where Drew adopts the assumption that the exposure decision the photographer would make by observing

•the camera histogram for a test shot taken with the camera WB set to some arbitrary setting

would most often be a lesser exposure than he would adopt by observing

•the camera histogram for a test shot taken with the camera's white balance set to a custom setting (CWB) derived by measurement of the shot environment (as with a calibration shot with a white balance measurement diffuser, or a calibration shot of a neutral test target).

That is not necessarily true. This of course depends on:

• What arbitrary WB correction vector was the camera set to, and

• What would be the WB vector correction vector of the "appropriate" CWB setting, and

• What was the chromaticity distribution in the scene?

I would like to point out that the histogram shown on the camera LCD, based on whatever setting, may still have more room to the right even when it abuts the right. We need to check that with the new cameras.

The reality, as so aptly and eloquently expressed by Andrew Rodney, is that if we aspire to squeeze out as much "to the rightness" for each scene situation as possible, we need to have an in camera three-channel raw histogram.

Best regards,

Doug

[Asher Kelman]

Originally Posted by Andrew Rodney

And I submit, if you use this information as a guide to the Raw data, you're fooling yourself. "More appropriate" could be said to be "less a lie" or "less incorrect". But this is the improper use of the tool, that's the bottom line. Its the improper understanding of what the Histogram is showing you based on the data you will end up with. Its a delusional, make yourself feel better button. If you're shooting Raw, if your goal is to ETTR or have some actual idea of the effect of exposure on Raw data, looking at an output referred, gamma corrected histogram, based on in camera JPEG processing is far from accurate, useful and a good use of that visual feedback. Producing a metadata tag that alters that histogram and has no effect on the Raw isn't useful. And worse, to write a piece supposedly exposing myths to sell a product, while in effect producing a myth in the process is inexcusable.

So Andrew, Bart, Doug, Mike or anyone else, just for for RAW, to get the most benefit of the unique emissive features of the LCD:

What is the least delusional setting so as to get some better clue of the lighting where in one color might be less represented in the data rich rich part of the histogram?

Can we just set files for RAW only, or do we still get the jpg delusions on the LCD?

Asher

[Doug Kerr]

Hi, Asher,

Can we just set files for RAW only, or do we still get the jpg delusions on the LCD?

Even if the camera is set for raw output only, it still develops an RGB image (not really good to think of it it as JPEG) for the on-camera display, and applies the white balance color correction vector dictated by the current WB setting in generating it.

I would hope that the WB setting that would make the on-camera display best represent what we would eventually get in a developed image would be the one nearest the WB correction we would use for actual development of the image. But maybe not.

Of course one way to do that is to make a WB determination just as if we were going to actually do the WB color correction in-camera. (I hate to say it, but this is a corollary of Drew's recommendation, applied to a situation where the real delivered image will come from external development of the raw data.)

But it in fact may be hopeless with today's camera displays.

Best regards,

Doug

[Andrew Rodney]

Originally Posted by Asher Kelman

What is the least delusional setting so as to get some better clue of the lighting where in one color might be less represented in the data rich rich part of the histogram?

Can we just set files for RAW only, or do we still get the jpg delusions on the LCD?

I don't think there's any setting even remotely close to what the Raw histogram would really show. The on camera settings might help a tiny, tiny bit one way or the other, but in the grand scheme of things, its pretty small. The current histogram is showing what the gamma corrected image would look like after being baked by the camera processor which is pretty darn different from a Raw, linear encoded histogram.

[Mike Shimwell]

Doug et al

Here's a thread from Luminous Lanscape on this topic - Gluik is pretty good on this stuff I think.

Mike

[Doug Kerr]

Hi, Mike,

Sorry I missed the link before.

Very interesting thread. (I still need to read more of it carefully.)

Another plus is that it brought to my attention PhotoBola's Raw Image Analysis tool (Rawnalyze). That is way neat, and very helpful in the areas we have been discussing.

Among other things, one could conceivably take a test shot, load the file into the 'puter, look at the raw histogram with Rawnalyze, and make an exposure decision. Not awfully handy, but doable. What do you think, Andrew?

Thanks again.

Best regards,

Doug

[Asher Kelman]

Originally Posted by Doug_Kerr

Hi, Mike,

Sorry I missed the link before.

Very interesting thread. (I still need to read more of it carefully.)

Another plus is that it brought to my attention PhotoBola's Raw Image Analysis tool (Rawnalyze). That is way neat, and very helpful in the areas we have been discussing.

Among other things, one could conceivably take a test shot, load the file into the 'puter, look at the raw histogram with Rawnalyze, and make an exposure decision. Not awfully handy, but doable. What do you think, Andrew?

Anyone know a Mac version?

[Andrew Rodney]

Originally Posted by Doug_Kerr

Not awfully handy, but doable. What do you think, Andrew?

I've never attempted Gluik's technique but on paper (and with his illustrations and scientific logic), it seems quite sensible unlike what Drew proposes and can't back up. I agree, it seems like a lot of work in some situations like shooting landscapes on location with a laptop, but the concepts seem totally sound to me.

[Bart_van_der_Wolf]

Originally Posted by Mike Shimwell

Doug et al

Here's a thread from Luminous Lanscape on this topic - Gluik is pretty good on this stuff I think.

Mike

Hi Mike,

Yes that's what I also had in mind.

Camera whitebalance settings will only influence the post capture colorbalance of the (preview) JPEG, and thus the histogram. While changing the in camera WB might help a little in avoiding a too conservative exposure (by avoiding too early clipping warnings), it's not a real substitute for an actual light filter.

The optimal physical exposure of the Raw image can benefit from a magenta filter in daylight conditions. Under tungsten lighting the blue channel can use some more help, red is already boosed by the illuminant.

Bart