Subjective WB Question

Mike J · May 1, 2009

So...here are four variants of the same shot taken in overcast skies (what else do we have here in Oregon at this time of year?). Only difference in this the WB setting in LR. I've gone back and forth as to what I think subjectively looks best. Which do you prefer? Would also be interested in by the book methods as well...

#1 As shot by the camera (D300) with WB set to auto. LR temp reads 4450. Tint is 0 for all of the shots.

#2 Auto adjust in LR. Changes temp to 3600

#3 LR custom adjusted to temp of 5200

#4 LR custom adjusted to temp of 5500

Elaine · May 1, 2009

I personally prefer #3, with #2 as my least favorite. #1 doesn't look bad at first, especially next to #2, but #3 is nicely warm and makes #1 look a bit cool, IMO, while #4 is a touch warmer than needed. I could see the argument for going somewhere between 1 and 3 as well.
There's my .02!

AFBlue · May 1, 2009

Mike J wrote:

So...here are four variants of the same shot taken in overcast skies (what else do we have here in Oregon at this time of year?). Only difference in this the WB setting in LR. I've gone back and forth as to what I think subjectively looks best. Which do you prefer? Would also be interested in by the book methods as well...

I like #3 best, but the differences among the custom setting examples does not look all that significant to me.

the Pixel Doctor · May 1, 2009

#3. And I (personally) would selectively darken the background to minimize it's distraction and promote the subject.

Dennis
aka: the Pixel Doctor

Psalm 150:6

cmason · May 1, 2009

Why not sample the white logo on her shirt and use that as your custom WB?

rutt · May 1, 2009

The issue here is that there is no obvious known neutral to balance on. The letters on her vest are some sort of high visibility reflective thing and might not be pure white. The man's shirt and cap in the right background might be navy. Ditto her sleeves.

Her flesh reads too magenta vs yellow (or in RGB, too green vs blue.) So that's a clue. When that happens we'll want to compare with a version with flesh more in line with the rule of thumb that flesh should be at least as yellow as magenta.

What about the black on the baseball glove? That might actually be black and not a very dark brown.

I tried balancing on this and liked the results; the flesh also came closer to the rule of thumb. I tried balancing the flesh to be Y>=M and like those results too. In both cases the greenery in the background was more yellow (no surprise) which might not be quite as nice. You can darken this or tone down the yellow in various ways if it matters to you. IMHO, better to get nice fleshtones than nice vegetation if you have to choose.

This doesn't have to be a completely subjective decision. See this thread.

arodney · May 1, 2009

Anything but #2. #1 seems OK to me, but since you're the artist, you decide if you want to convey a warmer look such that #3 would be fine with me, the viewer.

We have no visual clue as to the time of day. So is it noon or late in the day? And do you want to convey this (is it important)? There's no right answer here, but I'd suggest #2 is the wrong one. Is there anything wrong with shooting under high noon and producing an image that appears to be later in the day, when the light is warmer (answer for me, nope).

Candid Arts · May 1, 2009

I say #1. #2 is WAY to cool, and #3 is a little to warm, while #4 is way to warm.

I don't think #1 is quite there yet, but I'd use that as a starting point and play with the WB slider, possibly camera profile (at least that's what it's called for my and my canon gear, don't know how much affect it has on Nikon stuff, but it's changed colors dramatically (sometimes for the better) for me) as well as the tint slider.

Good luck.

Mike J · May 1, 2009

Thanks everyone. I appreciate your opinions here. It is pretty much a consensus that #2 is the wrong one. When I do this purely subjectively, I lean toward the warmth in #3 or #4 but when I compare what I like to what the camera does (#1,) I wonder if I lean too far towards the red. Hence- the reason for asking other's opinions.

cmason wrote:

Why not sample the white logo on her shirt and use that as your custom WB?

This is what I did as a starting point to get the 5500 temp. Sampling the white logo on the front gets a temp of ~5500 (last photo). Sometimes a tad lower. Sampling the white on her shoulder (more reflective?) puts it way up to 6900.

rutt wrote:

This doesn't have to be a completely subjective decision. See this thread.

Been following your other threads on this subject so I appreciate you commenting here. I'm still learning the unsubjective methods...

rutt wrote:

Her flesh reads too magenta vs yellow (or in RGB, too green vs blue.) So that's a clue. When that happens we'll want to compare with a version with flesh more in line with the rule of thumb that flesh should be at least as yellow as magenta.

So...let's restrict ourselves to doing this in LR where I can just get RGB readouts in %. In your other threads, I believe it is suggested that the G be a little higher than blue. Here are the RGB values (taken from her cheek) of #3 and #4

#3 - R:84%; G:76%; B:69%
#4 - R:84%; G:76%; B:67%

Do you have any quantitive suggestions for skin tones?

What about the black on the baseball glove? That might actually be black and not a very dark brown.

In LR, this yields things in ~5200 - 5700 range with no changes to the tint.

arodney wrote:

We have no visual clue as to the time of day. So is it noon or late in the day? And do you want to convey this (is it important)? There's no right answer here, but I'd suggest #2 is the wrong one.

This was taken at 6:10PM which is about 2 hours before our sunset. As mentioned, it was in high-overcast skies. It started to pour down rain just 1/2 hour later.

Candid Arts wrote:

I don't think #1 is quite there yet, but I'd use that as a starting point and play with the WB slider, possibly camera profile (at least that's what it's called for my and my canon gear, don't know how much affect it has on Nikon stuff, but it's changed colors dramatically (sometimes for the better) for me) as well as the tint slider.

This is very similar to what I do. I generally start with the standard Nikon profile for the D300 and sometimes will go to the landscape or vivid profiles. I'll sample the photo/play with the temp slider to get something that looks good to me. Generally I don't play with the tint slider but I think I need to spend more time figuring that out.

rutt · May 1, 2009

Mike J wrote:

Been following your other threads on this subject so I appreciate you commenting here. I'm still learning the unsubjective methods...

So...let's restrict ourselves to doing this in LR where I can just get RGB readouts in %. In your other threads, I believe it is suggested that the G be a little higher than blue. Here are the RGB values (taken from her cheek) of #3 and #4

#3 - R:84%; G:76%; B:69%
#4 - R:84%; G:76%; B:67%

Do you have any quantitive suggestions for skin tones?

I guess I got it wrong. Not used to using RGB for this. Magenta is the opponent of green. Where there is magenta, there is no green and visa versa. So more magenta means less green. Similarly with yellow and blue. That means you can translate the rule of thumb from the CMYK space to RGB: you want at least as much green as blue and red should be substantially more than either green or blue, maybe 30% more. Measure some flesh tones in pictures you lkke and keep doing it. After a while you'll get the feel for it.

The point here is not to replace your eyes and artistic judgment, but to give yourself a foot on the ground so you avoid the mistake of being fooled by your eyes into thinking your shot could be improved. It's sort of like having a recipe or a meat thermometer for cooking. Sure the good cook tastes and improvises, but s/he has some simple structure to start with.

arodney · May 1, 2009

Mike J wrote:

#3 - R:84%; G:76%; B:69%
#4 - R:84%; G:76%; B:67%

The difference numerically are so tiny that no matter what "by the numbers for skin tone" guide you use, it falls between the cracks. Depending on the color space, 2 values colorimetrically are very tiny. What do YOU prefer?

There's no right answer here.

Do you have any quantitive suggestions for skin tones?

This is a big, scary and largely unnecessary rabbit hole you don't need to dive into. The same girl shot under noon overcast sky, non overcast sky, 2 hours before sun set, at sunset, the values are all different. Place a different person in the scene, different values. Just look at the huge differences in skin tone even among Caucasians on TV, in film or on the web. The right answer is the one you find you prefer. This is all output referred so there is no way to measure accuracy here*. You're working towards pleasing color. You're not trying to match a piece of artwork you shot on a display or even a print, a difficult task.

Ask 100 photographers which of the four they like and expect for #2 which looks off, I suspect you'll get all kinds of different answers that will not help you get to your goal unless you're editing your images for a committee or consensus.

*http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf

rutt · May 1, 2009

Andrew and I have a deep and old disagreement here. We are in danger of hijacking your thread. So let me just point you at this old thread. I'm not sure either of us has much to add other than that we still haven't changed our minds.

Mike J · May 1, 2009

arodney wrote:

There's no right answer here.

That's for certain and I knew that going into this...my point was to see if what I like lines up loosely with other opinions.

We did start to get a consensus (albeit with a small number of votes so far. #2 is too cool and people lean towards the warmer versions of #3 and #4. It turns out that my tastes line up with this. I processed the entire batch taken at this time in LR to a temp of 5200 - 5500 tweaking individual photos to account for different skin tones.

Andrew and I have a deep and old disagreement here. We are in danger of hijacking your thread. So let me just point you at this old thread. I'm not sure either of us has much to add other than that we still haven't changed our minds.

I've followed your disagreements on this subject over time and no you are not in danger of hijacking this thread. I got I what wanted from both of you - your opinion on the first 4 photos. As Andrew said, there is no right answer. I'm just after opinions. I also find value in comparing my tastes to the "by the numbers" approach as it is another anchor for me to double-check against.

My aproach to WB is pretty unscientific - try a few of the presets in LR to see if there is anything I like. After that, I try to set the WB by sampling a couple of points in the photo. Between those approaches, I usually find something that looks right to me.

After settling on that value, I do a quick check of the RGB values to make sure that nothing is really out of whack. As you said, you do get a sense of what the values "should" be after doing this to a variety of photos.

I appreciate everyone's input and hopefully we can get a few more to chime in...:D

jfriend · May 2, 2009

Here's some non-subjective input based on numbers.

In image #1, I put sample points in four places to sample the color values, two on her face, one on the white lettering on the uniform and one on her dark shirt right above the red vneck. I then switched the color sampler for the two points on the face to show me CMYK (SWOP) values. Some people can do this in RGB, but I find it easier to see the relationships in CMYK (SWOP).

In Image #1:

The two CMYK (SWOP) facial values are:

C 9%
M 22%
Y 13%
K 0%

and

C 8%
M 19%
Y 10%
K 0%

We usually want to see Yellow be just a little more than Magenta for typical caucassian skin tone. If Yellow is below Magenta, then the image is generally not warm enough (e.g. it's too blue) for normal light. If you want to convey a shot in the shade, the image is sometimes supposed to be a little bluer, but that is not usually the most pleasing.

Then I look at the sample point on the lettering. I agree with Rutt than we don't know for sure that it's white, but it's worth a look. In Image #1, it's 248, 254, 254 which is low on red to be white. Nothing to conclude for sure with this one.

Then, I look at the sample point on the dark shirt. It reads 44, 58, 95. Wow, that's got a lot of blue in it for a dark shirt. It doesn't look like it's an actual blue shirt and blue-black is kind of unusual. It seems more likely it's actually supposed to be closer to black. And, if it were that would be completely consistent with the facial tones having too much blue in them. Removing blue would both put the facial tones where we expect them and make the shirt be more black. That looks like a promising guess.

If I then look at the black on her glove, I see 21, 32, 50. Since it's again unlikely this is a blue black, this is another clue that there's too much blue.

If I sample the blurred gentleman's shirt/jacket in the background, it shows 11, 15, 24. Again, probably too much blue. This would be an awful lot of coincidence to have so many blue blacks in the scene.

If I sample the little bit of her pants, it's 12, 15, 22. More blue black.

OK, I think it's safe to conclude that odds are really good that there's too much blue unless you want it to look like a heavily shaded scene.

If we then look at image #3, we see two facial sample points in CMYK (SWOP) as:

C 6%
M 26%
Y 26%
K 0%

and

C 3%
M 21%
Y 22%
K 0%

We have Yellow equal to or a little more than Magenta. That's what we expect.

The lettering on the uniform is 253, 253, 253 - perfectly neutral.

The shirt is 65, 63, 84. That's still on the blue side. It's less blue than image #1. If we want to make it more black, we could by using a curve to only adjust the darks. But, let's look at the other darks first.

The blurred gentleman in the background's black top shows 16, 17, 19. That's very close to neutral.

The black on the glove shows 28, 29, 34. That's pretty close to neutral.

Her pants are 14, 14, 16. Again very close to neutral.

I would conclude that image #3 is pretty darn close to reality. It could be shown bluer if you want to portray shade. It could be shown warmer if you want to portray the last rays of sunset light. Of the four, my preference is for image #3, but you could tweak it slightly depending upon what look you want to portray.

I personally think the darker tones in image #3 are a little too light and the image gets a little more pop if you add the bottom half of an s-curve (something that pulls the bottom tones down a little bit without really affecting the upper tones.

rutt · May 2, 2009

Mike J wrote:

I also find value in comparing my tastes to the "by the numbers" approach as it is another anchor for me to double-check against.

Exactly what I think is the right thing to do.

John's post also showed a great example of the kind of by-the-numbers reasoning which I think is appropriate when there is no obvious neutral to use for balance. I'll bet his results and mine would be virtually identical, perhaps an indicator that we are doing something not subjective after all. In fact, we could each have done it with a B&W monitor. (Though whether the result would be what you want might be subjective.) The claim of the by-the-numbers approach is that even if it isn't what you want in any particular case, on average, for most images, for a large majority of people, the result will be preferred when it conforms to a few rules of thumb (flesh, sky, obvious neutrals, water, animal fur not blue.) So if you are going to break these rules:

You should be aware that you are doing so, and
You should know why.

arodney · May 2, 2009

jfriend wrote:

I would conclude that image #3 is pretty darn close to reality.

Who's reality? Really, that kind of discussion is boarding on silly. In reality, #2 is "closer" as the OP stated it was shot under cool, overcast light. The only way to really know what "reality" is would be to have taken a spectral reading of the light that day and the girls skin under that illuminant. And then examine the Raw scene referred data and attempt to reconstruct it output referred to produce a match. And the OP would have disregarded this as not appearing attractive. The entire discussion of reality or correct or accurate is BS unless you can measure the colors at the scene and reconstruct it output referred. We want pleasing color. So #3 is more pleasing? That's just an opinion as this is totally subjective.

The right answer is the one the creator prefers. Unless he's a politician and trying to produce images by committee.

As to the CMYK values, using what sampling size? 1x1, 3x3, 100x100? The numbers are the same? Really, some of you are making this far more difficult than necessary for the OP. And what's the final output? You think it will appear the same on an sRGB (hopefully calibrated display in an ICC aware application of which many here are not using to view the images in a browser)? A printer? Which one? Talk about splitting hairs numerically. I'm all for numbers. The numbers a Spectrophotometer provides of the scene for the task of producing accurate color is damn useful even if the ultimate goal is an image no one likes. Arguing about a few CMYK values, values that have no bearing whatsoever on the content or output of the image is useful for those who are more impressed with number than images.

See what I mean Mike? You really want to go down that rabbit hole, or continue to make images?

rutt · May 2, 2009

Here is the disconnect. The question is whether to try and eliminate the effects of lighting or not and if so whether it can be done "objectively". Andrew would say that it cannot be done objectively and John and I would probably say that it can be, at least a little. Does the human visual system do something when viewing a scene which it does not do when viewing a picture of the scene on the monitor or a print? I think it does and that makes it desirable to try to correct the image to capture that experience. I suppose that Andrew also believes this, but I'm not quite sure. What I'm sure he doesn't believe is that there are a few qualitative rules which capture most people's preferences for most images most of the time. John and I do believe it.

Fine. No reason for WWIII. The techniques are useful to you if they are and not otherwise. Nevertheless, I do think it's a mistake to reject them before learning and trying them because they have proven themselves useful to many people over time.

jfriend · May 2, 2009

arodney wrote:

Who's reality? Really, that kind of discussion is boarding on silly. In reality, #2 is "closer" as the OP stated it was shot under cool, overcast light. The only way to really know what "reality" is would be to have taken a spectral reading of the light that day and the girls skin under that illuminant. And then examine the Raw scene referred data and attempt to reconstruct it output referred to produce a match. And the OP would have disregarded this as not appearing attractive. The entire discussion of reality or correct or accurate is BS unless you can measure the colors at the scene and reconstruct it output referred. We want pleasing color. So #3 is more pleasing? That's just an opinion as this is totally subjective.

The right answer is the one the creator prefers. Unless he's a politician and trying to produce images by committee.

As to the CMYK values, using what sampling size? 1x1, 3x3, 100x100? The numbers are the same? Really, some of you are making this far more difficult than necessary for the OP. And what's the final output? You think it will appear the same on an sRGB (hopefully calibrated display in an ICC aware application of which many here are not using to view the images in a browser)? A printer? Which one? Talk about splitting hairs numerically. I'm all for numbers. The numbers a Spectrophotometer provides of the scene for the task of producing accurate color is damn useful even if the ultimate goal is an image no one likes. Arguing about a few CMYK values, values that have no bearing whatsoever on the content or output of the image is useful for those who are more impressed with number than images.

See what I mean Mike? You really want to go down that rabbit hole, or continue to make images?

Andrew, if you believe that you should ONLY ever use your eyes and should never look for numerical clues as to what should and shouldn't be in the image, then that is your prerogative. I'm fine with you teaching other methods for adjusting images if you care to. You just come across to us here as unhelpful in the discussion if all you want to do is argue against these techniques without offering an alternative that would be more useful to everyone reading the thread.

The fact is that everyone does not have a highly trained eye for what "looks good" and Rutt and I are tyring to offer some tools that can be helpful when your eyes are confused or unsure. When I'm confused about how an image really should be, I find it very useful to look around and see what clues we can find in the image that guide me to things that probably are not supposed to be true or to confirm that this is probably right. They are not absolutes. I will never leave an image with the numbers how I like them, but an undesirable looking result. But, that usually doesn't happen. The clues point me in the right direction. They either confirm that what I seem to like is probably the best it can look or they point me in a direction that makes an image I like better and they gave me some guidance about what direction to go to get there.

When I say "reality", I'm talking about the color something really is in neutral light. It is highly unlikely that all those blacks in that image actually have that bluish component to them. It's possible that the shirt, pants and jacket in the background are all part of a uniform kit and could all have that in them. But, it would be highly unusual to see that same bluish tint in the glove too. My daughter actually owns that glove and I've never thought of it as blue at all.

I find it useful to get this information. The image is probably showing a blue cast. If you have that as a starting point, it's easy to then say to yourself that that's how you think it should show because of the lighting or to move a little warmer or a little cooler to get the appropriate lighting effect. If you want some guidance on how much to move to get rid of the blue cast, you can look at the numbers while you adjust the image and try values around a neutral cast to see if that gives you something you like. In the end, I always stop on something my eyes like, but these are tools to help me get there, make sure I'm not being fooled and make sure that if there is a cast, it's there on purpose.

FYI, I don't do this on very many images. I do it when I'm confused, when simple and typical adjustments don't quickly lead to what I like or am expecting or when I can't really decide how the image should look and I'm in search of some other data to help guide me. And, clearly you have to watch out for things that can confuse the issue even further like mixed lighting (which this image probably does not have). But, its a tool that I find helpful in some circumstances.

rutt · May 2, 2009

I guess my eyes aren't as good as John's. I measure almost every image and often this leads me to make improvements I wouldn't have seen otherwise. When I measure images posted on dgrin (especially in the people forum) I often find the same thing. In addition to magenta skin, I've also found blue squirrels and horses, green skies, and red tuxedos. In each case nobody was seeing it. I mean nobody. Instead, the question was often, "Does this look better in color or B&W." Or "How can I make this pop?" Or "This looks great on my calibrated monitor but terrible in print." So people knew there was something wrong, but they seeing what exactly.

So I measure always. Often it stops there. But often I learn something useful. Then I can ignore what I've learned or not. I think this is like knowing good grammar. Once you master it, you may find that you can be more expressive when you break the rules. But you break the rules unknowingly at your peril.

pathfinder · May 2, 2009

This is an interesting discussion of the philosophy of color correction, that so far, has remained civil, and informative. I find things of value in both sides of this discussion.

I would like to know from the original poster if the shirt and the pants from the young girl's uniform were perceived at the scene, as blue or as black. They are certainly blue in the first image, and beginning to approach black in the last image.

I tend to favor somewhere between #1 and #3, but I suspect the original poster knows whether the shirt and pants were seen as black or as blue on the field. That fact might influence me a great deal.

I do know that I do not feel compelled to match the color temperature the scene was shot under in my final edited image. I agree that using known neutral colors and known colors ( as read as pixel date ) can be quite helpful in color correcting images, but I also know that my final images are, sometimes, deliberately far from this when finally printed, as a deliberate artistic choice, as I think Andrew was suggesting. Sometimes an image looks better warmer than shot - this is frequently true of people and sunsets among other things.

It is always good to know several different ways to achieve a goal as an artist.

Mike J · May 2, 2009

pathfinder wrote:

This is an interesting discussion of the philosophy of color correction, that so far, has remained civil, and informative. I find things of value in both sides of this discussion.

I would like to know from the original poster if the shirt and the pants from the young girl's uniform were perceived at the scene, as blue or as black. They are certainly blue in the first image, and beginning to approach black in the last image.

I tend to favor somewhere between #1 and #3, but I suspect the original poster knows whether the shirt and pants were seen as black or as blue on the field. That fact might influence me a great deal.

I do know that I do not feel compelled to match the color temperature the scene was shot under in my final edited image. I agree that using known neutral colors and known colors ( as read as pixel date ) can be quite helpful in color correcting images, but I also know that my final images are, sometimes, deliberately far from this when finally printed, as a deliberate artistic choice, as I think Andrew was suggesting. Sometimes an image looks better warmer than shot - this is frequently true of people and sunsets among other things.

It is always good to know several different ways to achieve a goal as an artist.

I could not have said this any better. I like learning the techniques as a sanity check but in the end, I make 'em like I like 'em.

I honestly don't recall with certaintity the preceived color of her shirt and pants but, as John stated, I would be suprised if they were bluish. The girls are told to wear black underneath their jerseys and I don't remember anyone not doing this.

I appreciate the philosphy of color correction that everyone has put forth. Just like you and I can differ about our opinions on a photo, I think there will always be differences here (technique, taste, etc).

Now...I note that John does his checking in CMYK. For those of us limited to LR and PSE (doesn't do CMYK), what are the equivalent checks in RGB?

arodney · May 2, 2009

Even if we could agree that our memory color of the scene were true and accurate (which is easy to dismiss), we are not digital capture or output devices that only handle what we call color using sets of numbers. Color is a phenomenon that happens in our brains. So the question becomes, are the numbers representative of the scene? We can measure the colors and the illuminant but that entire useage of color science isn't at all as robust as necessary because we don't have color appearance models in current software applications. They instead are all based on science done in the 1930's long before anyone could imagine what a computer was, let alone Photoshop or Lightoom. This is why we experience optical illusions like this among many others:http://web.mit.edu/persci/people/adelson/checkershadow_illusion.html

The color models are at fault here (Lab, the color space so many feel is so darn powerful has many such mathematical frailties that cause all kinds of issues like blues shifting magenta).

We also adapt to white, capture devices don't. There's a host of other such disconnects we could go into, but the point is, we can't say its objective unless we have something concrete to base this on (unless you guys want to go by your gut feelings in which case, count me out of the discussion, I don't care to go into religious debates). We can discuss objective measurements as I tried to allude to above. Otherwise, its totally subjective.

This isn't to say numbers are not useful. I've been reading numbers in Photoshop since 1990, using version 1.0.7 (same with Histograms). I've probably measured more colors and produced more deltaE reports based on number than most. Numbers can be useful, they can back up what our visual systems attempt to tell us but there's a point where they are not useful and shouldn't be used in the context of images to tell us something we see and don't agree with. Lets look at some examples of where objectiveness really isn't objective:

If I then look at the black on her glove, I see 21, 32, 50. Since it's again unlikely this is a blue black, this is another clue that there's too much blue.

If I sample the blurred gentleman's shirt/jacket in the background, it shows 11, 15, 24. Again, probably too much blue. This would be an awful lot of coincidence to have so many blue blacks in the scene.

If I sample the little bit of her pants, it's 12, 15, 22. More blue black.

OK, I think it's safe to conclude that odds are really good that there's too much blue unless you want it to look like a heavily shaded scene.
The lettering on the uniform is 253, 253, 253 - perfectly neutral.

The shirt is 65, 63, 84. That's still on the blue side. It's less blue than image #1. If we want to make it more black, we could by using a curve to only adjust the darks. But, let's look at the other darks first.

The blurred gentleman in the background's black top shows 16, 17, 19. That's very close to neutral.

The black on the glove shows 28, 29, 34. That's pretty close to neutral.

Her pants are 14, 14, 16. Again very close to neutral.

I would conclude that image #3 is pretty darn close to reality.

Not to pick on the post per say, this goes back to the fellow who has suggested we do a lot of these corrections based on numbers. He often uses the same tactic "it should be this or it could be this" etc "...unless its not". This color guru doesn't like saying anything that could be written in stone which gives the by the numbers model a pretty big fudge factor. Well should and could don't cut it objectively, or at least scientifically. The image above was shot under cool light. We either match that exactly or we subjectively decide what we want. That adjacent colors could "pollute" the spot we are reading numerically, or the numeric values don't produce a desired color appearance which should be kept in mind here.

You can say the glove is supposed to be black but the reality is, we don't know unless we measure it and we see the effect of the color under an illuminant (note that without light, there's no color so its kind of important yes?).

Its a big guessing game and, it dismisses facts we don't have the actual measurements of any or all of the above items, under the captured illuminant. And it dismisses that even if we did, we might all very well not like the color appearance.

You could have shot this girl, holding a gray card under sunset light, white or gray balanced it and totally removed the warm effect on any or all of the above items mentioned. Would this make a better image or one that's "truer" to the scene? In fact, would have moving the tint/temp slider warmer improved the image as we see in this example of a sunset from Martin Evening's tutorial on WB?:

http://lightroom-news.com/2009/04/27/mastering-the-white-balance-controls/

If you want to for go objective numbers, well you better damn measure them! And the numbers will change based on the media used to view them. Otherwise this all becomes a case of mental masturbation. In such a case, no one can agree what is numerically correct or accurate. So again, I'm all for numbers. I use them all the time. I use them to back up what I'm doing visually when I render images. But there's a point where everyone has to accept its subjective. You can tell someone "numerically that image isn't correct" and if you do, prove it! The only way I know to prove it is with a device that can measure the numbers, otherwise, its a pointless debate.

Use numbers, use your eyes, use your aesthetics but don't let numbers alone tell you your eyes or aesthetics are wrong. If this were true, we'd all have excellent auto routines in our software products, our output devices etc. Sometimes such routines work pretty well, sometimes they fail totally. Sometimes they get us closer but we have to tweak.

arodney · May 2, 2009

Mike J wrote:

Now...I note that John does his checking in CMYK. For those of us limited to LR and PSE (doesn't do CMYK), what are the equivalent checks in RGB?

Its far easier in LR:

http://digitaldog.net/files/LR_Skintone_Ratio.jpg

YMMV, but IF you must use ratio's you can see that you only have to deal with three and they are roughly (like all the other by the numbers should be correct for skin values), 10% apart.

arodney · May 2, 2009

rutt wrote:

Here is the disconnect. The question is whether to try and eliminate the effects of lighting or not and if so whether it can be done "objectively". Andrew would say that it cannot be done objectively and John and I would probably say that it can be, at least a little.

I suppose that Andrew also believes this, but I'm not quite sure. What I'm sure he doesn't believe is that there are a few qualitative rules which capture most people's preferences for most images most of the time.

What I believe is, the numbers that produce a color appearance you desire are the correct numbers and are neither objective or accurate, because we'd have to use a non human, non biased instrument to prove the point. I no more feel the above image is "correct" if the black is neutral than I do that a black dress at sunset should be neutral. Pleasing color produces a huge set of numbers we view in context that is "correct" (quotes should be emphasize!) if you, the image creator prefer those numbers.

jfriend · May 2, 2009

arodney wrote:

In such a case, no one can agree what is numerically correct or accurate. So again, I'm all for numbers. I use them all the time. I use them to back up what I'm doing visually when I render images. But there's a point where everyone has to accept its subjective. You can tell someone "numerically that image isn't correct" and if you do, prove it! The only way I know to prove it is with a device that can measure the numbers, otherwise, its a pointless debate.

Use numbers, use your eyes, use your aesthetics but don't let numbers alone tell you your eyes or aesthetics are wrong. If this were true, we'd all have excellent auto routines in our software products, our output devices etc. Sometimes such routines work pretty well, sometimes they fail totally. Sometimes they get us closer but we have to tweak.

The part I quoted above from your post is in exact agreement with what I said in my post. I said to use the numbers to give you clues for whether your eyes are being fooled or which way you might try adjusting the image and in the end go with what looks good to you. The numbers are not magic. They are not the "truth" that should be blindly followed. They are tools to consider, just like the histogram is a tool on your camera to help with exposure. You don't blindly follow it, but you use it for guidance. I do agree with Rutt that when an image goes counter to the normal rules for the numbers, it's useful to know you're doing that and double check that you're sure this is what you want to do.

The skill for adjusting an image is as much about knowing which tools you can use to get to something you like as it is in having good judgment about what looks right or good. I'm just trying to share how others can use these tools to provide more info when adjusting or evaluating images.

If I read your post right, you are arguing that blindly following numbers is bad, but they can be valuable tools. If that's what you meant, then we are in agreement.

rutt · May 2, 2009

Violent agreement.

rutt · May 2, 2009

If you really want to find out more than you ever thought you'd want to know on this subject, read this old thread. I'm not sure that either Andrew or I have added anything. John did figure out that Andrew agrees with us, which might be a breakthrough. (Though we seem to have gotten there in the aforementioned thread, but it wasn't said well enough to stick.)

Subjective WB Question

Comments