Independent White Balance Control In LR

[smoke665]

It is an old thread. Since then I can confirm that the FOSS app DarkTable is capable of setting multiple WB multipliers for different regions of a photo as is DXO PL. It's a option in PL that I use regularly now.

This was a quote from another old thread so rather than continue it, I'm starting a new one. Your comment prompted me to look at LR, now that they've had all the updates, to see how this might work.
This is the original image.

And this is the edited image in LR, using selective masking (color range) to alter the white balance (only) in the background, the leaf, and the stems. I went a little overboard for illustration purposes. There many other selection options that could also be used to target more specific areas, from object selection, luminosity, depth range, even AI options like ground cover, pavement, sky, background, etc

Knowing your preference for non destructive editing, this was completed with 4 masks, each of which can be edited or even completely deleted, without affecting the original image.

How does this compare to what you're using???

 

[Ysarex94 88]

This was a quote from another old thread so rather than continue it, I'm starting a new one. Your comment prompted me to look at LR, now that they've had all the updates, to see how this might work.
This is the original image.
View attachment 292511
And this is the edited image in LR, using selective masking (color range) to alter the white balance (only) in the background, the leaf, and the stems. I went a little overboard for illustration purposes. There many other selection options that could also be used to target more specific areas, from object selection, luminosity, depth range, even AI options like ground cover, pavement, sky, background, etc
View attachment 292512
Knowing your preference for non destructive editing, this was completed with 4 masks, each of which can be edited or even completely deleted, without affecting the original image.

How does this compare to what you're using???

Did LR allow you to mask a region of the photo and change the WB or did it only allow you to change the color. In other words assume the original image has a WB set to 5000K tint = 0. Does LR allow you to mask an area of the image and then change 5000K to 6550K -- can you type in a new WB temp value for the masked region?

As I think you observed in the old thread for a photo like a landscape or the photo here the difference between the two procedures isn't going to be a big deal. We could illustrate how the two procedures differ with a color checker, but it will be an academic exercise.

So here's the color checker test. The original file is a raw CR2 file from a Canon camera.

Image A is entirely white balanced in C1 from sampling the middle grey square on the color checker. Color for the pillow, canister, couch and color checker are good. Temp value is 4606K

I then set the WB by using the sampler on the Caucasian skin tone square, 2nd down from upper right corner, and got the color in image B for the pillow, canister and couch. Temp value is 3214K. But I masked the color checker and set the WB for the masked area from the middle grey square again (4606K). Color checkers match image A and B.

I then loaded the same raw file in Affinity. Affinity has a WB adjustment layer that doesn't really set a WB -- it tries to white balance on top of an existing WB. I started with the raw file and set the WB with the sampler from the Caucasian skin tone square. Color in image C of the pillow, canister and couch are similar to image B. I then masked the color checker and used the WB adjustment layer to sample the middle grey square. That's a pretty whacked color checker. The degree of whacked will depend on the degree K discrepancy and the software used, but I've never seen any photo processing app that tries to set a WB on top of an existing WB ever be able to get an accurate color checker in the process.

What's going to happen out in the natural world in a landscape is that you'll be adjusting the color to something you like and odds are you won't have a color checker to worry about. In a cityscape with brightly colored cars, signs, people's cloths, etc. trying to set a WB on top of an existing WB will likely generate some whacky colors but odds are you won't notice as long as what you're focusing on becomes the color you like.

 

[smoke665]

Did LR allow you to mask a region of the photo and change the WB or did it only allow you to change the color. In other words assume the original image has a WB set to 5000K tint = 0. Does LR allow you to mask an area of the image and then change 5000K to 6550K -- can you type in a new WB temp value for the masked region?

As I think you observed in the old thread for a photo like a landscape or the photo here the difference between the two procedures isn't going to be a big deal. We could illustrate how the two procedures differ with a color checker, but it will be an academic exercise.

Lightroom allows you to enter temperature as kelvins in the base panel. When you apply a mask it pulls up a new basic panel that includes Temperature, tint and hue as "sliders". Adjustments made here are on a sliding scale 0-100 (+or-) to the base temperature setting. As you said, not a big deal if they accomplish the same end game. Just interesting how things work.

 

[Ysarex94 88]

Lightroom allows you to enter temperature as kelvins in the base panel. When you apply a mask it pulls up a new basic panel that includes Temperature, tint and hue as "sliders". Adjustments made here are on a sliding scale 0-100 (+or-) to the base temperature setting. As you said, not a big deal if they accomplish the same end game. Just interesting how things work.

That's how I remember both LR and PS. And from that I surmise what they're doing is adjusting color on top of an existing WB. If you like the color as adjusted fine, but if you try the color checker exercise like I did above you won't be able to get an accurate color checker from a neutral WB sample.

Most photo processing apps only permit 1 WB in the image. And of course if you start with an RGB processed image like a camera JPEG then the WB is already cooked in and it can't be changed. The place I run into this issue the most is when I encounter folks who edit camera JPEGs insisting that they can change the WB -- they can't, they can change colors and it's not the same thing.

 

[smoke665]

@Ysarex using your methodology I tried to duplicate it in Lr. This first image is the original. I set the WB to the gray box 2nd from the bottom left at 5800k

I then went in and using color range, to mask the dress, I changed the Temperature slider only to -63, and brushed the color checker out of the mask.

As you can see it significantly altered the hue of the dress, but the color checker box still reads 5800. If you apply a mask selection, and make adjustments it affects the areas revealed, there may or may not be bleed through where the mask doesn't completely cover other things, that's why the mask has the option to add or subtract areas. Luminosity difference in the image will affect not only WB, but selection, as will similarities in hues when using color range, making manual masking adjustments necessary.

WB adjustment is nothing more than a R,G,B multiplier, it doesn't sound logical to me that the software would alter that multiplier every time you changed an area of the image rather than using a single stacking adjustment, especially in a non-destructive environment. It would create exponential instructions to remember. I couldn't find anything to confirm or disclaim your assertion. Do you have anything???

 

[Ysarex94 88]

@Ysarex using your methodology I tried to duplicate it in Lr. This first image is the original. I set the WB to the gray box 2nd from the bottom left at 5800k
View attachment 292516
I then went in and using color range, to mask the dress, I changed the Temperature slider only to -63, and brushed the color checker out of the mask.
View attachment 292517
As you can see it significantly altered the hue of the dress, but the color checker box still reads 5800.

You've made no change to the color checker. You began by setting a WB from the color checker. You then masked the dress and made sure the color checker was not masked so the change you made changed the dress. WB remained the same and the area masked was changed. That's not trying to do what I did in Affinity that caused the color checker error.

1. Shoot a raw file that includes the color checker and some other colored items.
2. Open the raw file and set the WB from the grey square as you did by using the WB sampler. Save for reference.
3. Now reset the WB by using the sampler on one of the colored squares -- 2nd square down from the upper right corner. You're going to get a whacky WB.
4. Mask the color checker and try and adjust the color sliders until the same grey square you originally sampled is grey, or I don't know if LR will permit using the WB sampler for a masked area but try that. Compare the adjusted masked color checker with the first one you saved.

Or try this step 4. Send the image to PS and once in PS open the Camera Raw Filter and try the white balance sampler on that grey square -- compare color checkers.

If you apply a mask selection, and make adjustments it affects the areas revealed, there may or may not be bleed through where the mask doesn't completely cover other things, that's why the mask has the option to add or subtract areas. Luminosity difference in the image will affect not only WB, but selection, as will similarities in hues when using color range, making manual masking adjustments necessary.

WB adjustment is nothing more than a R,G,B multiplier, it doesn't sound logical to me that the software would alter that multiplier every time you changed an area of the image rather than using a single stacking adjustment, especially in a non-destructive environment. It would create exponential instructions to remember. I couldn't find anything to confirm or disclaim your assertion. Do you have anything???

Only lots and lots and lots of tests over many years.

Here's ACDSee's attempt. Much better than Affinity but still a complete failure. Just look at the greyscale in image B. I don't think the square just below white is very light grey.

Photo A is white balanced for reference from middle grey. Photo B was white balanced from 2nd square down from top right. Then the color checker was masked and the white balance sampler was used to set WB for the color checker. It's an attempt to set WB on top of an existing WB and it crashes and burns.

 

[Ysarex94 88]

Here's my antique copy of PS (CS6) making the attempt. Much better than the other two -- I'm impressed. But still a failure; the white square of the greyscale values =
R 223
G 213
B 204
Twenty point spread between red and blue is a crash.
I opened the raw file in ACR and set the WB (sampler) from the 2nd square down from the top right. Sent the image to PS and used the Camera Raw Filter to set the WB (sampler) from the middle grey square (color checker masked). This has to be the hundred plus time I've done a similar version of this with lots of different apps. Try and set a WB on top of an existing WB and if the discrepancy is at all substantial you'll never get an accurate color checker or a grey greyscale.

 

[Ysarex94 88]

I asked Perplexity and yep, makes sense. Even in a raw processor like LR you're looking at an RGB white balanced and tone curve adjusted image. And if LR is adjusting that data in a second attempt to set a WB then the answer below applies.

"Adjusting white balance on an RGB photo that already has a white balance baked in works poorly because the data has already been rescaled once, and you are trying to “rebalance” numbers that no longer represent the original scene in a simple, linear way.

What white balance does​

White balance isn’t just a cosmetic filter; it is a channel‑by‑channel gain applied to the sensor’s red, green, and blue responses to make a neutral object have equal RGB values under a particular illuminant. In a raw workflow this is done before converting to a standard color space, with high bit depth and access to the camera’s full dynamic range.

What happens in a baked RGB image​

Once the raw data is rendered to an 8‑bit RGB image (JPEG, sRGB TIFF, etc.), the white balance chosen at that time is “baked in” together with tone curves, clipping, and compression. Significant color shifts at this stage require stretching or compressing already‑quantized channel values, which exaggerates rounding errors, banding, and noise, especially in shadows and saturated areas.

Why a second white balance is worse​

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.

Limits of “neutral picker” tools on RGB​

Eyedropper/neutral‑picker tools assume that if a pixel should be neutral, its three channels just need to be equalized to fix the whole image. In a previously white‑balanced RGB file, that assumption breaks down when different tonal regions have been affected differently by the first rendering (tone curves, local contrast, clipping), so forcing one area to neutral can easily introduce color errors elsewhere.

Why raw behaves differently​

With raw data, the converter can change white balance non‑destructively by altering per‑channel gains before tone mapping in a high‑bit‑depth space, so you can make large corrections with relatively little damage. With an already‑processed RGB image, you are limited to small tweaks around the existing balance; beyond that, artifacts and color inaccuracy become unavoidable because the original scene information has been irreversibly compressed and clipped."

 

[smoke665]

asked Perplexity and yep, makes sense. Even in a raw processor like LR you're looking at an RGB white balanced and tone curve adjusted image. And if LR is adjusting that data in a second attempt to set a WB then the answer below applies.

Okay you've shifted direction on me, in the original statement you said "capable of setting multiple WB multipliers for different regions of a photo". Now if I understand correctly you're saying changing the WB on the complete raw image. To further clarify LR does not modify or change the WB in the raw file. LR "creates" a 1:1 JPEG preview of the raw file. ALL instructions apply to the preview only. The raw file isn't even brought into LR, it remains in its original folder as pristine as ever.

So my understanding is there's only two methods of editing, destructive (making actual changes to a raw file) or nondestructive (not changes to the raw file). To work nondestructive, both LR and C1 create a JPEG preview for editing while the raw file remains untouched. So each would "hard bake" the WB per your explanation. So what's different? What am i missing?

Once the raw data is rendered to an 8‑bit RGB image (JPEG, sRGB TIFF, etc.), the white balance chosen at that time is “baked in” together with tone curves, clipping, and compression.

I haven't worked 8-bit in a long time. LR runs as a 64-bit application that works with image data in 8, 16, and 32-bit. PS can handle up to 32-bit (with limitations). I normally work with 16-bit in both. I dont use the file formats you referenced until the final export.

altering per‑channel gains before tone mapping in a high‑bit‑depth space

What's considered "high-bit-depth space? "With an already‑processed RGB image, you are limited to small tweaks around the existing balance", as pointed out above, both C1 & LR use a "processed" RGB preview? Whats different?

 

[TJMcG121 4]

Ysarex said:

Once the raw data is rendered to an 8‑bit RGB image (JPEG, sRGB TIFF, etc.), the white balance chosen at that time is “baked in” together with tone curves, clipping, and compression.

Good discussion of color temp and color. I've never understood why some people think that they can edit an 8-bit JPEG the same as a 12-14-bit raw file.

 

[smoke665]

Good discussion of color temp and color. I've never understood why some people think that they can edit an 8-bit JPEG the same as a 12-14-bit raw file.

I'm a "why" type of person that likes to understand the inner workings of something. Joe is a good source of knowledge, that makes me question and think!

 

[Ysarex94 88]

Okay you've shifted direction on me, in the original statement you said "capable of setting multiple WB multipliers for different regions of a photo". Now if I understand correctly you're saying changing the WB on the complete raw image.

No change -- that's what I meant. C1 and PL can both apply multiple WB multipliers for different regions of a photo. LR can't.

To further clarify LR does not modify or change the WB in the raw file. LR "creates" a 1:1 JPEG preview of the raw file. ALL instructions apply to the preview only. The raw file isn't even brought into LR, it remains in its original folder as pristine as ever.

All raw processors work with raw data and then create an RGB preview image for you. Instructions can apply to the raw data and those instructions produce the RGB preview. Or some instructions secondarily are applied to the RGB data. That's what's happening here. PL/C1 will apply WB multipliers first to all the raw data but with the option to select/mask a separate region and apply different WB multipliers to only the selected/masked region. Both sets of WB multipliers are applied selectively to the raw data to create the RGB preview. LR will apply WB multipliers to all the raw data with no option to apply different WB multipliers to a selected/masked region. Instead LR will alter color in the RGB preview on top of the already applied global WB.

So my understanding is there's only two methods of editing, destructive (making actual changes to a raw file) or nondestructive (not changes to the raw file). To work nondestructive, both LR and C1 create a JPEG preview for editing while the raw file remains untouched. So each would "hard bake" the WB per your explanation. So what's different? What am i missing?


I haven't worked 8-bit in a long time. LR runs as a 64-bit application that works with image data in 8, 16, and 32-bit. PS can handle up to 32-bit (with limitations). I normally work with 16-bit in both. I dont use the file formats you referenced until the final export.

That was just Perplexity making some assumptions -- I cut and pasted the entire answer -- the AI assumed someone trying to edit a JPEG. The same situation occurs editing a raw file and working with 16 bit RGB data but likely with less extreme whackiness in the results (more room to push and pull).

What's considered "high-bit-depth space? "With an already‑processed RGB image, you are limited to small tweaks around the existing balance", as pointed out above, both C1 & LR use a "processed" RGB preview? Whats different?

There's some blame to hand out here. Adobe is fine, there engineers have made a choice -- you get one global WB applied to the raw data and after that you can adjust color. They still give you the temp/tint sliders but the scale changes and you can no longer set a degree K value -- you're not resetting WB. If I'm editing a landscape I can handle that just fine, I want to make a color change for example to the sky -- no problem -- I can mask the sky and make the change. The blame goes to some of the other software vendors like Serif/Canva and ACDSee who put tools into their software that they name WB and which include a WB sample tool. Folks get the idea that those tools change WB and they really don't; they change color on top of the existing WB and that produces different results than setting new WB multipliers to raw data.

 

[smoke665]

No change -- that's what I meant. C1 and PL can both apply multiple WB multipliers for different regions of a photo. LR can't.

Might have been your intent, but I didn't read it that way.

No change -- that's what I meant. C1 and PL can both apply multiple WB multipliers for different regions of a photo. LR can't.

All raw processors work with raw data and then create an RGB preview image for you. Instructions can apply to the raw data and those instructions produce the RGB preview. Or some instructions secondarily are applied to the RGB data. That's what's happening here. PL/C1 will apply WB multipliers first to all the raw data but with the option to select/mask a separate region and apply different WB multipliers to only the selected/masked region. Both sets of WB multipliers are applied selectively to the raw data to create the RGB preview. LR will apply WB multipliers to all the raw data with no option to apply different WB multipliers to a selected/masked region. Instead LR will alter color in the RGB preview on top of the already applied global WB.

Okay so if I understand it correctly, you're saying that after the intial WB setting, C1 uses raw data to make changes to the WB (global or selective) of the preview, while LR makes +/- adjustments to the preview without consulting the raw data? I'll need to read up on C1 but initial research didn't reveal that. Do you have anything that says that?

 

[Ysarex94 88]

Might have been your intent, but I didn't read it that way.

Okay so if I understand it correctly, you're saying that after the intial WB setting, C1 uses raw data to make changes to the WB (global or selective) of the preview, while LR makes +/- adjustments to the preview without consulting the raw data? I'll need to read up on C1 but initial research didn't reveal that. Do you have anything that says that?

How about an example: That's my neighborhood's main street. We're at the city's highest point and so we can see the Arch.

The scene is back/side lit; sun is shining at me from the right. Left side of the street is sunlit as is the sky. That's daylight WB and I set daylight WB for the entire image in the background layer. Below the image are two of C1's control panel tools, Layers and White Balance. In the set where the Background Layer is selected (orange) the WB values are temp = 5650, tint = 3.5. That was originally applied to the entire image.

With daylight WB set for the entire image the right side of the street is in open shade and it's as blue as the sky and I don't like those extreme blue shadows. So I created a luminosity mask for the shadows and used the WB sampler to click on that parked white car which was a blue car when I clicked on it. In the second set of control panel tools you see that Adjustment Layer 2 is selected (that's my mask) and WB values are temp = 6884, tint = 3.9. C1 went to the raw data and for only the masked region it used new/different WB multipliers than were used for the entire image set in the background layer.

In LR you can create a luminosity mask just as I did in C1 and you can warm up the shadows to remove the blue. LR will change color on top of the daylight WB and the tool tells you that by not allowing you to set a new degree K value, whereas C1 gave me the option to set two different sets of WB multipliers to selected regions of the raw data. Unless you're running a test with a color checker you're not likely to see any difference.

 

[smoke665]

How about an example:

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??????