Independent White Balance Control In LR

[Ysarex96 90]

You can show examples all day, the issue isn't "if" it's different, but why is it different????

The engineers at C1 and DXO decided it was a valuable feature and the engineers at Adobe didn't.

I posed the question on a couple different forums including Adobe, curious to see the responses. For now the only thing I can find is LR & C1 "have different starting points and workflows due to their proprietary RAW processing engines and default color profiles". They both use masks to apply subsequent adjustments, I can't find anything to elaborate on "how" the process works on those adjustments....is it stacked...is it a color multiplier... does it use the raw data??????

It has to be using the raw data -- the color checker test works and that proves it.

 

[smoke665]

It has to be using the raw data -- the color checker test works and that proves it.

Not necessarily. I went back and tried your suggestion in LR. The image on the left is the original...WB set to 5400k. The one on the left I raised the WB to 7500k (note the option to change WB to a specific kelvin remains as an option in the Basic Panel), then masked the color checker and reduced the WB -54 (quess) to what looked to be similar to the original, because the masking function only allows slider +/- (not able to set specific kelvin). Point color readings on the gray box (2nd up from btm left) on the original was RGB 80,77,85 reading on the edited copy same square is RGB 76,77,87. There's a difference yes but hardly "wanky", and as I said I was guessing by eye. I could have gone back and refined it more to get it closer. Having the ability to set temperture to K in masking would be quicker no doubt, but I seriously doubt that most people would be able to tell the difference here on the CC card.

 

[TJMcG122 6]

You can show examples all day, the issue isn't "if" it's different, but why is it different???? I posed the question on a couple different forums including Adobe, curious to see the responses. For now the only thing I can find is LR & C1 "have different starting points and workflows due to their proprietary RAW processing engines and default color profiles". They both use masks to apply subsequent adjustments, I can't find anything to elaborate on "how" the process works on those adjustments....is it stacked...is it a color multiplier... does it use the raw data??????

I asked Google's Gemini, and here is what it said:

:

1. Capture One: True Layered Adjustments​

C1 is structured to work more like a simplified Photoshop than Lightroom.

• Stacking Method: C1's local adjustments are built on a true layer system. You create a new layer (with a mask) and on that layer, you have access to a full suite of tools, including the White Balance tool itself.
• "How" it Works: When you apply a WB adjustment on a C1 layer, that adjustment is stacked on top of the global RAW WB and is applied only to the pixels under the mask. C1's engine is sophisticated enough that many of these localized adjustments (including WB) are still calculated using the RAW data's dynamic range, similar to the global adjustments. The result is a clean, precise local WB correction.

2. Lightroom: Masked Adjustments/Brushes​

Lightroom's local tools are fundamentally different from C1's layers.

• Stacking Method: LR applies a set of plus/minus adjustments (Exposure $+, 0.5$, Contrast $-10$, etc.) to a masked area. While you have Temperature and Tint sliders in the local adjustment panel, these are technically color-shifting adjustments applied on top of the global WB.
• "How" it Works: While Adobe claims these local adjustments still benefit from the RAW data's larger dynamic range, they are conceptually post-global adjustments. The process is likely stacked in a complex way:
  
  1. Global WB (Multiplier on RAW data)
  2. Local Masking (Applying a color/tone shift to a masked section of the rendered image data, still using the RAW data's wide color space as the source).

In both cases, they are applying a form of White Balance Multiplier (or a complex color/tone shift that emulates a multiplier) to the masked area, but C1's method is more explicit and gives the user full access to the WB tool within a layer, leading to its reputation for superior local color control.

 

[smoke665]

@TJMcG thank you for this ^^^^. As I read, it appears that "both LR and C1" stack the adjustment mask on the underlying global WB. There might well be some discrepancy in the terminology used (WB adjustment vs Color/tone shift), but the end result is still basically the same. C1 has a supposedly superior color process, but I suspect that all the recent updates to LR have narrowed that gap, as the illustration I just posted above shows. I do wish LR had the ability to set kelvin vs using the slider.

 

[Ysarex96 90]

Not necessarily. I went back and tried your suggestion in LR. The image on the left is the original...WB set to 5400k. The one on the left I raised the WB to 7500k (note the option to change WB to a specific kelvin remains as an option in the Basic Panel), then masked the color checker and reduced the WB -54 (quess) to what looked to be similar to the original, because the masking function only allows slider +/- (not able to set specific kelvin). Point color readings on the gray box (2nd up from btm left) on the original was RGB 80,77,85 reading on the edited copy same square is RGB 76,77,87. There's a difference yes but hardly "wanky", and as I said I was guessing by eye. I could have gone back and refined it more to get it closer. Having the ability to set temperture to K in masking would be quicker no doubt, but I seriously doubt that most people would be able to tell the difference here on the CC card.

Doing that exercise in C1 and you get two identical color checkers -- all squares match exactly and they should because you're setting the same WB multipliers for the same raw data.
The less of a degree K and tint discrepancy the closer you'll get in LR but you have to compare all squares of the color checker. Doing it by eye is too painstaking. You do have enough difference in the two that if you put the squares side by side you can see it in the Hue values -- the numbers I put inside the squares. Some are closer one to the other and some are farther apart which is a real bad sign because it means there's no global correction available.

View attachment 292558

 

[Ysarex96 90]

@TJMcG thank you for this ^^^^. As I read, it appears that "both LR and C1" stack the adjustment mask on the underlying global WB. There might well be some discrepancy in the terminology used (WB adjustment vs Color/tone shift), but the end result is still basically the same.

Not necessarily. Try the color checker test the way I did it by setting the WB with the sampler from the color checker square 2nd down from top right. That's a substantial discrepancy. Then mask the color checker and try and get it back to a clean greyscale and the right colors in all the squares. Even if you manage to get the same grey square in both versions to match you won't have all matching color squares.

Not sure about the forum's conversion of images so here's a link to the image below.

Dropbox

www.dropbox.com

You can download it and load it into PS and compare color squares between the two color checkers -- they are all identical Hue values. You can't do the same in LR, it won't work.

C1 has a supposedly superior color process, but I suspect that all the recent updates to LR have narrowed that gap, as the illustration I just posted above shows. I do wish LR had the ability to set kelvin vs using the slider.

 

[Ysarex96 90]

Got a correction here: Because DXO PL allows a selected region to be masked and will then let you set degree K values for WB for that region I assumed it was working the same as C1. But I had never done the color checker test to verify that. I just did the test and it fails. It does not set a second and separate WB with the raw data. It changes color on top of the base WB and even though it accepts what appears to be new degree K values, PL can not set a second and separate WB for a selected/masked region.

DarkTable is the only other app I know of that supposedly will do what C1 does -- guess I'll have to test that as well. Could be that C1 is unique.

 

[TJMcG122 6]

Got a correction here: Because DXO PL allows a selected region to be masked and will then let you set degree K values for WB for that region I assumed it was working the same as C1. But I had never done the color checker test to verify that. I just did the test and it fails. It does not set a second and separate WB with the raw data. It changes color on top of the base WB and even though it accepts what appears to be new degree K values, PL can not set a second and separate WB for a selected/masked region.

DarkTable is the only other app I know of that supposedly will do what C1 does -- guess I'll have to test that as well. Could be that C1 is unique.

Why is this important? Who needs exact Kelvin temps in edits?

 

[smoke665]

ou do have enough difference in the two that if you put the squares side by side you can see it in the Hue values

Through out I've never said there weren't differences. Different software there's bound to be. I said the differences weren't "wanky" as you stated.

As stated before my question was why. TJ provided the best answer, its the processing engine and the algorithms. Unlike your earlier comment it seems that both stack color, and both use raw data. The difference is in how its interpreted.

I've done a lot of color grading by the numbers over the years. Taking samples from any given area can read different, especially colors, because luminosity and saturation can change the hue drastically. I used the gray box because it's the least affected. Still if your sample size is small you cant get an accurate reading. I used a 31x31 sample, and still it can vary. The only way to be semi accurate is to select an area, average and then sample.

 

[smoke665]

Why is this important? Who needs exact Kelvin temps in edits?

Agreed! I use HSB valves to color match when tone mapping in PS, though i must say that LR with their AI powered adaptive profiles are pretty awesome.

 

[Ysarex96 90]

Through out I've never said there weren't differences.

In C1 there are no differences between the two color checkers -- they are identical. And that's because C1 works differently than LR and most other apps. C1 really can set two different white balances using the raw data. LR can not and so LR can not make the two color checkers match if the WB discrepancy is substantial. I wouldn't switch from LR to C1 for that reason.

Different software there's bound to be. I said the differences weren't "wanky" as you stated.

As stated before my question was why. TJ provided the best answer, its the processing engine and the algorithms. Unlike your earlier comment it seems that both stack color, and both use raw data.

If they use raw data it's indirect. From the answer TJ found: "these are technically color-shifting adjustments applied on top of the global WB." Which is what I've always said they do. C1 does not do that -- C1 works differently, it replaces the global WB for the selected area.
The info I got from Perplexity correctly applies to what LR is doing and the results confirm that: "Applying another white‑balance correction later means applying new multipliers on top of the old ones rather than replacing them. This double scaling can push some channels into clipping, distort hues (skin tones, skies), and shift neutrals imperfectly because the mapping from scene colors to current RGB numbers is no longer a simple “illumination plus neutral” situation but a compound transform."
You can't even do the color checker test in LR exactly as I did because you can't sample WB for the masked color checker and you have to eyeball it. If you do the test as I did and start with a WB from the Caucasian skin tone and then finally succeed in getting a neutral grey square in the masked color checker you won't have matching color squares and some of the discrepancies will be pretty big.
Do this in LR:
Set WB with the sampler from the Caucasian skin tone square. Then pass a smart Object to PS. All the raw data is embedded in the smart object. In PS mask the color checker on the smart object layer and then bring up the Camera Raw Filter and use the WB sampler on the middle grey square of the greyscale. That at least let's you use the WB sampler. The resulting color checker will not match a color checker from the same file with WB set once from a sampled grey square.

The difference is in how its interpreted.

I've done a lot of color grading by the numbers over the years. Taking samples from any given area can read different, especially colors, because luminosity and saturation can change the hue drastically. I used the gray box because it's the least affected. Still if your sample size is small you cant get an accurate reading. I used a 31x31 sample, and still it can vary.

I used an 11x11 sample to test your two color checkers and there was never more than a one point variation in multiple tries on the same square.

The only way to be semi accurate is to select an area, average and then sample.

 

[Ysarex96 90]

Why is this important? Who needs exact Kelvin temps in edits?

Yep, Smoke and I have both noted it's no big deal. If I had to abandon C1 and start using LR again this would be the least of my concerns.

 

[smoke665]

Yep, Smoke and I have both noted it's no big deal. If I had to abandon C1 and start using LR again this would be the least of my concerns.

Two MO mules arguing over a moot point.....yup sounds about right. happy trails till the next one.

 

[smoke665]

@Ysarex I wanted to get back to you with what I learned from Adobe community. A raw file has no white balance yet (technically a raw file is greyscale, because it only has one value per pixel rather than three), so the white balance needs to be set upon raw conversion. The 'as shot' value is a metadata entry that is read by Lightroom as a starting point, that conversion process is a proprietary algorithm (same as C1). In LR a mask is like a layer, a white balance adjustment in a mask is a color shift that is stacked on top of the base image. I am assuming that C1 works in a similar manner, but there are difference in how the metadata is processed.

In LR masked adjustments are relative to the image-wide settings. A "mask adjustment" does not set the masked area to an absolute value, it "offsets" the value of the underlying pixels by a slider amount, relative to each image’s current "image-wide setting". Adobe designed Photoshop/Lightroom/Camera Raw masks to work in the way that traditional photographers work. In the darkroom, a straight print is whatever the values are across the negative. When using a mask to dodge or burn, apply contrast or color filtration in the enlarger, the result is a shift relative to the original negative value.

Maybe this helps explain the difference?

 

[Ysarex96 90]

@Ysarex I wanted to get back to you with what I learned from Adobe community. A raw file has no white balance yet (technically a raw file is greyscale, because it only has one value per pixel rather than three), so the white balance needs to be set upon raw conversion. The 'as shot' value is a metadata entry that is read by Lightroom as a starting point, that conversion process is a proprietary algorithm (same as C1). In LR a mask is like a layer, a white balance adjustment in a mask is a color shift that is stacked on top of the base image. I am assuming that C1 works in a similar manner, but there are difference in how the metadata is processed.

In LR masked adjustments are relative to the image-wide settings. A "mask adjustment" does not set the masked area to an absolute value, it "offsets" the value of the underlying pixels by a slider amount, relative to each image’s current "image-wide setting". Adobe designed Photoshop/Lightroom/Camera Raw masks to work in the way that traditional photographers work. In the darkroom, a straight print is whatever the values are across the negative. When using a mask to dodge or burn, apply contrast or color filtration in the enlarger, the result is a shift relative to the original negative value.

Maybe this helps explain the difference?

Right -- it's what I've always said. LR let's you set one global WB for the image. You can change it but only globally for the entire image. Any adjustment of a masked area is applied on top of that globally set WB. C1 is different. It also set's a global WB for the entire image. But if you mask an area on a layer C1 will allow you to replace the global WB values for the masked area and apply a new WB for just the masked area. It doesn't make adjustments on top of the existing global WB but removes that global WB (masked area) and replaces it.