P: Strange white balance for overlayed NEF

[drtonyb]

Many Nikon DSLR's have an overlay function which combines (overlays) two NEF images to produce a third NEF image in-camera. The source images don't have to be different, so an image can be combined with itself, which is what I did for this example.

I have noticed that for such combined images, the WB for As Shot has Temp: 2880 and Tint: -119, regardless of the WB camera setting of the original images. In this example, I shot an image with my D750's WB set to Daylight. Importing the original image into Lr Classic 8.0 gives the As Shot WB as Temp: 5300 Tint: +7, which is always the value for this WB setting. See the screen-shot below:



If I change the WB from As Shot to Daylight, the values change slightly to Temp: 5500 Tint: +10, as shown:



Combining this image with itself using 0.5 for the contribution of both source images to make a new overlayed NEF and importing into Lr, in develop I see this:



White balance for the image looks fine, but the settings aren't right for the WB camera setting. Changing the WB from As Shot to Daylight shows how incorrect Lr has got the WB:



For comparison, Nikon's Capture NX-D gives the 'As Shot' WB settings Temp: 5209 Tint: 0 for both the original and the combined (overlayed) images. Changing settings to Direct Sunlight changes Temp: 5200 Tint: 0 in both cases without drastically changing the images' appearance.

The overlay camera function takes its settings from the first image selected for the overlay. So if overlaying two different images that were shot with different WB settings, the combined image has the WB settings for the first image. Capture NX-D reflects this - first image Daylight, second image Shade, the combined image has Temp: 5209 Tint: 0. First image Shade, second image Daylight, the combined image has Temp: 8006 Tint: 0. For Lr, either combined image has Temp: 2880 Tint: -119.

[drtonyb]

This post seems to have gone unnoticed by everyone!

[drtonyb]

Lr Classic 8.4 and 9 months later, this problem still hasn't been fixed, let alone being acknowledged. Extremely disappointed in Adobe's inability to fix bugs reported by its users.

A correction to my original post is that the default Temp is set to 2800, not 2880.

My understanding is that Lr evaluates its own WB value because it can't use Nikon's propriety methods. So how can Lr get this so wrong?

[Pavan Kumar T S]

Hello Anthony, I tried to reproduce this issue but don't have a D750 right now, so used a D800. I shot an image with manual exposure and WB set to daylight. Then overlay - same image in both the sections with 0.5 values in both and saved the image. Imported the single image and overlay image as well, and in both of them WB was 5,050 and +8. Changing it to daylight in LR set it to 5500. Please let us know if there are any other steps that you used or if that was all.

It would be great if you could share the raw file where this happens too. Thanks!

[drtonyb]

Hello Pavan, I have added two images to Dropbox for you to test. The original, DSC_0387.NEF was shot with camera WB set to daylight and aperture preferred exposure. The second image, DSC_0388.NEF is the overlayed image created in the same way you described, i.e. selected the original in both sections with 0.5 settings and saved.

In Lightroom Classic 8.4, after import the original image DSC_0387.NEF has an initial WB 'As Shot' Temp: 5300 Tint: +7 and when I change the WB setting to 'Daylight', I get Temp: 5500 Tint: +10 and the image still looks normal.

After import, the overlayed image, DSC_0388.NEF, has an initial WB 'As Shot' Temp: 2800 Tint: -119. Changing the WB setting in Lr to 'Daylight' gives the same as with the original image, Temp: 5500 Tint: +10, but the image has a horrible pink cast. Using the White Balance Selector will return the image to normal appearance when a 'grey point' is sampled, but the Temp and Tint are still wildly wrong. For example, zooming in on the beach in the image and sampling the darkest part of the sand gives me Temp: 2950 Tint: -128.

Files are here:

https://www.dropbox.com/sh/xpy0gcm0or5sypg/AAAYL5b-xMKpvhnfLaP41Dx7a?dl=0

Let me know if there is any problem downloading the files. I hope this helps.

Tony

[Pavan Kumar T S]

Thanks for the files Tony! I did try it and yes, the WB shows as 2800 and -119. Does that happen with all files or only this particular one? 

[drtonyb]

Hello Pavan,

This happens with all files, and also regardless of the original's WB setting.

Tony

[drtonyb]

Hello Pavan,

I've been looking into this a bit further and found that opening the overlayed image in RawTherapee gave much the same result. This made me think the problem is in the NEF, not in Lightroom. Checking the EXIF metadata I see that WB_RBLevels is "1 1 1 1" in the overlay image, but is "1.94921875 1.34375 1 1" in the original file.

My conclusion is that the D750 overlay function is applying a WB correction to each NEF file's data before combining them. This makes a bit of sense to me so that it can handle two images shot with different WB settings. It is a pity that the overlay function doesn't check to see if the two images have the same WB setting and then combine them without changing the red/blue values.

Tony

[drtonyb]

I have posted about this strange problem before and have done some more testing and investigating, but I'm still puzzled as to why LrC is getting the white balance so wrong for 'Overlayed' images from my Nikon D750.

This is a long post and many won't be very interested in reading it.

For those who may not know, Nikon DSLR's have an in-camera function that combines two RAW images to produce a new RAW image. This function is called 'Overlay' in which the user selects the amount or brightness of each image to contribute to the overlayed result. It's a bit like a double exposure function. There is nothing preventing the user 'Overlaying' an image with itself, as I have done here using a 0.5 amount setting in each to produce an Overlay image that should be the same as the original image.

I have taken one shot using my Nikon D750 and another with a D850, both with White Balance set to 'Direct sunlight', then used Overlay in each camera to produce a new RAW, resulting in four RAW images. Importing these into LrC and switching to the Develop module, the Temperature and Tint for each of the images are:

WB: As Shot
DSC_0142 (D750 original shot) Temp: 5300 Tint: +7
DSC_0143 (D750 overlay) Temp: 2800 Tint: -119
DSC_0179 (D850 original shot) Temp: 5250 Tint: +1
DSC_0180 (D850 overlay) Temp: 5500 Tint: +10



All images look normal with regards to their colour balance, but the Overlay image DSC_0143 from my D750 has a very weird WB Temperature and Tint as seen in the screen shot above.

Now I change the LrC WB setting on all images to Daylight. The resulting Temperature and Tint values are:

WB: Daylight
DSC_0142 (D750 original shot) Temp: 5500 Tint: +10
DSC_0143 (D750 overlay) Temp: 5500 Tint: +10
DSC_0179 (D850 original shot) Temp: 5500 Tint: +10
DSC_0180 (D850 overlay) Temp: 5500 Tint: +10

Three of the images look perfectly normal, but the Overlay image DSC_0143 produced by my D750 now has the 'right' Temp/Tint, but a horrid pink colour cast!



What I'm finding puzzling is that more recent Nikon Cameras than my D750, like the D850 and D810, don't seem to have a problem, not in LrC at least. I have tested with other RAW convertors like Nikon's Capture NX-D, RawTherapee and ON1 Photo Raw with mixed results. Capture NX-D gets the WB 'right' for all images, 'As Shot' or not, which is to be expected from Nikon's own software and their 'inside' knowledge. Raw Therapee and ON1 Photo Raw produce an 'incorrect' WB for both overlay images when set to 'As Shot' and a pink colour cast when both overlay images are set to 'Daylight'.

All this leaves me wondering how LrC can get the White Balance 'correct' for the D850 overlay image, but 'wrong' for the D750 overlay image.

Investigating deeper, I have found that the overlay images seem to have the white balance applied to the source image data before being combined. This makes sense if the source images are taken using different WB camera settings. In the Nikon Exif data there is a setting called WB_RBLevels which has four numerical values. My understanding is that these are multipliers for the Red Blue Green Green pixels in the RAW data and are applied to get the colour balance for the 'As Shot' setting. For the two Nikon cameras I'm using in my tests with the WB set to Direct Daylight, the original RAW images have multipliers:

D750 WB_RBLevels: 1.94921875 1.34375 1 1
D850 WB_RBLevels: 1.900390625 1.375488281 1 1

For the overlay images, the multipliers are:

D750 WB_RBLevels: 1 1 1 1
D850 WB_RBLevels: 1 1 1 1

The overlay images with multiplier values of 1 1 1 1 suggests to me that the WB multipliers have been applied to the RAW image data for the overlay images. Looking at the values in the RAW pixels (using RawTherapee without demosaicing), I can see that this appears to be the case by comparing Red, Blue and Green pixels in the overlay images to their values in the original source RAW images.

So how does a RAW convertor set the WB to something other than 'As Shot'? Simplistically, I can imagine that it just multiplies the RAW data by appropriate WB_RBLevels values chosen by the RAW converter. This would explain how the WB set to Daylight has a pink colour cast in the overlay images - the Daylight WB_RBLevels are effectively applied twice. Using ExifTool to change the WB_RBLevels of my original D750 RAW image to the square of the normal values:

D750 WB_RBLevels: 3.799453735 1.805664062 1 1

and then looking at the image in LrC with the WB set to As Shot, I see a result with the pink colour cast and Temperature 6750/Tint +150, which is what I expected. So, does LrC and other raw converters use the WB_RBLevels to work out the Temp/Tint for the As Shot WB? Is this how the values of 1 1 1 1 produce such incorrect values for As Shot Temp/Tint, except for Nikon's Capture NX-D?

What would happen if I set both original images WB_RBLevels to 1 1 1 1? I would expect a green colour cast for the As Shot WB and indeed, that is exactly what I get, for LrC at least. The D750 original image WB values for As Shot are now Temperature 2800/Tint -119, the same as its overlay image, while the D850 original's WB values for As Shot are Temperature 5500/Tint 10, also the same as its overlay image.

My investigation only partially explains things to me, but the question still remains. How can LrC get the As Shot WB 'correct' for the D850 overlay image when its WB_RBLevels are set to 1 1 1 1, while other RAW convertors are consistently 'incorrect' for this case? At least one thing is clear. LrC seems to be treating the WB for the D850 differently to the D750 when the WB_RBLevels are set to 1 1 1 1. For the D850, when WB_RBLevels is set to 1 1 1 1, LrC appears to be treating the WB as if Daylight was selected to determine the Temperature/Tint values - when WB_RBLevels = 1 1 1 1, WB As Shot and WB Daylight have the same Temp/Tint. However, for the D750 images, when WB_RBLevels = 1 1 1 1 the WB As Shot Temp/Tint values are very different to the WB Daylight values.

If you have been patient enough to get this far and have any light that you can shed on the problem, I would love to hear your comments.

[Todd Shaner]

Investigating deeper, I have found that the overlay images seem to have the white balance applied to the source image data before being combined.

The extreme WB difference appears to be happening in-camera when (as you state above) the WB is applied to the raw data. At the risk of stating the obvious have you had the D750 camera checked by Nikon Tech Support?

[drtonyb]

Hi Todd,

Thanks for your reply, although I believe the problem is not with the camera.

In the D750 overlay image, I changed the WB_RBLevels from 1 1 1 1 to the normal values for a Direct daylight WB camera setting. I expected this change to give me the 'correct' As Shot Temp/Tint and a pink colour cast in LrC. It does exactly that: Temp 5300 Tint +7, which is normal for the D750. For me, this rules out any fault with the camera.

When I do the same with the D850 overlay image, LrC displays the image with a pink colour cast and a WB As Shot Temp 5250 Tint +1. This is what I expected too. All other WB settings, like Daylight, also have the pink colour cast, which I also expected.

So my questions remain: when the WB_RBLevels is 1 1 1 1 in the D850 image, how or why is it treating its WB As Shot Temp/Tint differently to the D750 under the same WB_RBLevels setting? How is the D850 overlay NEF getting the correct WB without any pink colour cast for other WB settings, like Daylight, when its image data has already been multiplied by the WB_RBLevels for the Direct daylight setting in the camera? As I see it, this would require the data to be divided by the multipliers (which presumably LrC does not know the values for), then multiplied by the appropriate RB values for the selected LrC WB setting.

Could one of the development team please enlighten me?

[Todd Shaner]

It does appear that LR/ACR is processing the D750 image overlay WB differently than all other Nikon models. How the As Shot WB values are being determined for the image overlay file remains a mystery, but clearly Adobe knows how.

*Pavan Kumar you mentioned testing image overlay files shot with the Nikon D800 and no problem was found. You then mention testing *Anthony Blackett's D750 image overlay file DSC_0388.NEF and confirmed the WB values are incorrect (2800 and -119). Can you please forward this information to Adobe Engineering for review. Thank you.

[drtonyb]

Thanks again Todd for your input.

It's not just the D750, it also applies at least to the D90 overlay NEF files too, or as my tests showed, any NEF from these earlier Nikon DSLRs that has WB_RBLevels set to 1 1 1 1 (which is normally only the overlay files).

Looking forward to an answer, if Pavan or someone else from Adobe ever provides one.

[Todd Shaner]

Anthony, can you please clarify what you mean by, "any NEF from these earlier Nikon DSLRs that has WB_RBLevels set to 1 1 1 1 (which is normally only the overlay files)."


Are you saying that the D750 and earlier Nikon models have image overlay WB_RBLevels set to 1 1 1 1, but the D850 and later model image overlay files have normal WB_RBLevels (such as 1.94921875 1.34375 1 1)?

[drtonyb]

Hi Todd,

Are you saying that the D750 and earlier Nikon models have image overlay WB_RBLevels set to 1 1 1 1, but the D850 and later model image overlay files have normal WB_RBLevels (such as 1.94921875 1.34375 1 1)?

No, I'm not saying that. Overlay images from the D850 and other models later than the D750, such as the D810, also have WB_RBLevels set to 1 1 1 1 out of the camera and the raw pixel data has the WB_RBLevels from the original shots applied, just as the D750 does. Hence the WB_RBLevels are being normalised to  1 1 1 1 in the overlay images from all Nikon cameras.

What I am trying to say is that if I manually (Exiftool) set the WB_RBLevels value to 1 1 1 1 in the original raw files (not the overlay files), then in LrC for the D750 I see a WB As Shot Temperature 2800/Tint -119 and a green colour cast. This indicates to me that LrC is using the WB_RBLevels to apply the As Shot white balance and using it to calculate the Temp/Tint values.



If I select WB Daylight in LrC, then I get a Temperature 5500/Tint +10 and a normal looking white balance. In this case, I assume LrC is applying its own WB_RBLevels to the raw pixel data.



For the D850 original raw file with the WB_RBLevels manually set to 1 1 1 1, LrC gives a WB As Shot Temperature 5500/Tint +10 and the image has a green colour cast. What puzzles me here is where did LrC get the Temperature 5500/Tint +10 from? This is the same Temp/Tint values that LrC determines for its own WB Daylight setting, but somehow LrC has come up with these when it sees 1 1 1 1 for the WB_RBLevels in the exif data.



Now for an added twist, selecting the WB Daylight setting in LrC when the D850's WB_RBLevels is set to 1 1 1 1, I get a Temperature 5500/Tint +10, but the image has a green colour cast!



If I change the D850's WB_RBLevels to 0.999 1 1 1, LrC now reports WB As Shot Temperature 3100 /Tint -116 with a green colour cast.



Selecting WB Daylight then gives the Temperature 5500/Tint +10 and a normal looking image.

If I set the WB_RBLevels to 0.999 1 1 1 in the D750 original image, then I get the same results as if it were 1 1 1 1.

So, it appears to me that for a D850 image, having WB_RBLevels = 1 1 1 1 is triggering a different white balance algorithm in LrC than the one being used for the D750 with WB_RBLevels = 1 1 1 1.

Normally, the only images that will have WB_RBLevels = 1 1 1 1 are the overlay raw files from Nikon cameras and the algorithm used for the D850 appears to work for these images. If the same algorithm could be used for all Nikon cameras, including the D750, then I think the problem would be solved.

[Todd Shaner]

So in conclusion–I think this answers my last question.

Normally, the only images that will have WB_RBLevels = 1 1 1 1 are the overlay raw files from Nikon cameras and the algorithm used for the D850 appears to work for these images. If the same algorithm could be used for all Nikon cameras, including the D750, then I think the problem would be solved.

LR is not properly calculating the WB for image overlay files shot with D750 and earlier camera models. It is simply reading the the WB_RBLevels 1 1 1 1 value, which is Temp 2880 and Tint -119. The algorithm used for calculating WB in image overlay files for D850 and later models needs to be applied to the D750 and earlier model cameras.
I think that's pretty clear.

[Rikk Flohr: Photography]

It seems like I ask this question a lot but just to be clear: Is ACR behaving the same or differently as Lightroom Classic behaves on these files? 

[Todd Shaner]

Yes, ACR 12.3 exhibits the same incorrect As Shot WB settings with the D750 (and earlier models) image overlay NEf files.

[Rikk Flohr: Photography]

Super. the Camera Raw team is working on it now. Thanks for all the extra information!

[Todd Shaner]

Good to hear–Thanks Rikk!

[drtonyb]

Thanks Todd and Rikk. Just confirming Todd's reply that ACR is the same as LrC.

[Todd Shaner]

A similar issue with Canon Multi-Exposure CR2 files was reported at the below link.

https://feedback.photoshop.com/conversations/lightroom-classic/camera-rawlightroom-classic-incamera-...