DXO 2 & Calibrated Sigma 18-35mm

[Manfred M]

A question I asked earlier got a bit lost. Can anyone comment on the pros and cons of passing a dng or a TIFF from PhotoLab to another programme?

The difference is that the DNG file is nothing more than a raw file that is inside a wrapper. This simply means that when you export it to another program, the receiving program will treat it like the raw data file that it is. Colour space will not have been assigned nor will the white balance have been "baked in" nor with image bit depth be assigned. These value can be changed by the upstream raw converter without any issues that affect image quality. Gamma is also technically variable, but pretty well all computers and editing software use a 2.2 value, so this is not of great importance.

The TIFF is an image file, so color temperature, colour space, bit depth and gamma will all be "baked in". Sometimes these can be changed (colour temperature cannot), but there can be quality consequences. TIFF files support layers and support lossless compression. Photoshop PSD files are little more than a proprietary TIFF file with a few bells and whistles. TIFF files tend to be fairly large...

[Manfred M]

I don't understand why I do. And I'm not the only one I know.
https://support.dxo.com/hc/en-us/art...h-DxO-PhotoLab.

George

George I and the link you posted and have played around with this, and like Donald, I see no difference in image quality. The two methods use a different rendering approach. For smaller images, the rendering is computed by the DxO software. For larger images (75% of full size and larger) the work is sent to the computer's graphics card for rendering. In theory, the results should be equally good as the Open CL is nothing more than a tool that lets the software use the GPU directly, which is a lot faster than using the CPU.

If you are seeing a difference I would suspect that this might be related to your computer's graphics card and how well it runs Open CL.

That being said, this is only something that will affect the displayed image, rather than the image you are working on, so it can cause you to make incorrect edits based on incorrect output to screen. You can disable the use of Open CL in the PhotoLab settings. You might want to try this and see if that makes any difference in what you see displayed. Look under the Edit menu; Preferences selection and Performance tab.

[george013]

George I and the link you posted and have played around with this, and like Donald, I see no difference in image quality. The two methods use a different rendering approach. For smaller images, the rendering is computed by the DxO software. For larger images (75% of full size and larger) the work is sent to the computer's graphics card for rendering. In theory, the results should be equally good as the Open CL is nothing more than a tool that lets the software use the GPU directly, which is a lot faster than using the CPU.

If you are seeing a difference I would suspect that this might be related to your computer's graphics card and how well it runs Open CL.

That being said, this is only something that will affect the displayed image, rather than the image you are working on, so it can cause you to make incorrect edits based on incorrect output to screen. You can disable the use of Open CL in the PhotoLab settings. You might want to try this and see if that makes any difference in what you see displayed. Look under the Edit menu; Preferences selection and Performance tab.

I found the reason of this behaviour. When I open a Nef the standard DxO preset is used. It contains a level of sharpening. Below the 75% that sharpening is not computed, above it is. When I use the no-correction preset I don't see any difference.
I don't like this. I'm not able to view the image and its quality at a lower zoom as 75%. And another thing is, I can't resize the image within the program, only when exporting.
All things I could and can do within the Nikon software.

George

[xpatUSA]

<> Gamma is also technically variable, but pretty well all computers and editing software use a 2.2 value, so this is not of great importance.

<>

For most of the images posted on the 'net (sRGB) you probably know the actual gamma is 2.4 plus a straight bit:

If RL, GL, BL are less than or equal to 0.0031308
R = 12.92 RL
G = 12.92 GL
2
B = 12.92 BL
If RL, GL, BL are greater than 0.0031308
R = 1.055 RL(1/2.4) – 0.055
G = 1.055 GL(1/2.4) – 0.055
B = 1.055 BL(1/2.4) – 0.055

See www.color.org/srgb.pdf

Yes, I know that the above approximates to 2.2 gamma.

Still, it should also be noted that the popular ProPhoto working space (non-conforming Adobe versions excluded) has a gamma of 1.8, not 2.2.

See http://www.color.org/ROMMRGB.pdf

Of possibly equal importance, ProPhoto's white point is D50, not D65.

Excluding Adobe's variants, there are still plenty people using ProPhoto for editing which means that a good percentage of editing software/computers are using 1.8 gamma, whether they know it or not ...

[Manfred M]

Excluding Adobe's variants, there are still plenty people using ProPhoto for editing which means that a good percentage of editing software/computers are using 1.8 gamma, whether they know it or not ...

Which cannot be displayed on any computer screen I am aware of, so the output is going to be, at best, sRGB or Adobe RGB after the image is rendered in a colour space that can be displayed.

[davidedric]

Thanks, Manfred, and please pardon my ignorance.

Does that mean that the lens corrections from PhotoLab will not be passed over in a .dng file, but will be in a TIFF?

Dave

The difference is that the DNG file is nothing more than a raw file that is inside a wrapper. This simply means that when you export it to another program, the receiving program will treat it like the raw data file that it is. Colour space will not have been assigned nor will the white balance have been "baked in" nor with image bit depth be assigned. These value can be changed by the upstream raw converter without any issues that affect image quality. Gamma is also technically variable, but pretty well all computers and editing software use a 2.2 value, so this is not of great importance.

The TIFF is an image file, so color temperature, colour space, bit depth and gamma will all be "baked in". Sometimes these can be changed (colour temperature cannot), but there can be quality consequences. TIFF files support layers and support lossless compression. Photoshop PSD files are little more than a proprietary TIFF file with a few bells and whistles. TIFF files tend to be fairly large...

[Manfred M]

Thanks, Manfred, and please pardon my ignorance.

Does that mean that the lens corrections from PhotoLab will not be passed over in a .dng file, but will be in a TIFF?

Dave

No. It means that edit data will be written to the DNG file. The DNG is an Adobe open licensed product that encapsulates both the original raw file as well as the edit data (normally saved as metadata in a sidecar file), so any program that can open and read the data will be able to apply those edits.

I have not tested this extensively, but the lens correction data seems to come across cleanly as I get different results than when I open the raw data with Adobe software.

[xpatUSA]

ProPhoto cannot be displayed on any computer screen I am aware of, so the output is going to be, at best, sRGB or Adobe RGB after the image is rendered in a colour space that can be displayed.

Of course it can't, Manfred! Did you think I was unaware of that, or was the comment aimed at others?

Please explain why you said "Gamma is also technically variable, but pretty well all computers and editing software use a 2.2 value, so this is not of great importance." (my emphasis).

[Manfred M]

Of course it can't, Manfred! Did you think I was unaware of that, or was the comment aimed at others?

Aimed at others. Of course you know this, but not everyone does.

Please explain why you said "Gamma is also technically variable, but pretty well all computers and editing software use a 2.2 value, so this is not of great importance." (my emphasis).

As you have mentioned not everything uses a gamma of 2.2. For the longest time Apple hardware used a gamma of 1.8, but have been using 2.2 for the past decade or so. That being said the vast majority of users here will be on either PCs or Apple hardware which use a gamma of 2.2, hence while other values are possible it is not a value we can influence the settings. The popular software used in editing, especially in light of the Adobe / DxO discussions here has no way of letting the user change that. I only mentioned it only for completeness as the gamma value is "baked in" when the raw data is converted to an image file.

[xpatUSA]

Aimed at others. Of course you know this, but not everyone does.

As you have mentioned not everything uses a gamma of 2.2. For the longest time Apple hardware used a gamma of 1.8, but have been using 2.2 for the past decade or so. That being said the vast majority of users here will be on either PCs or Apple hardware which use a gamma of 2.2, hence while other values are possible it is not a value we can influence the settings. The popular software used in editing, especially in light of the Adobe / DxO discussions here has no way of letting the user change that. I only mentioned it only for completeness as the gamma value is "baked in" when the raw data is converted to an image file.

I get your drift in that only "the popular software" is worthy of mention here. It's a great pity that less popular (RawTherapee) software has ProPhoto as it's default working space and, beyond that, allows you to define and select any gamma you like in the output color space. For dabblers like myself, such a freedom to "influence the settings" is a delight, should it ever be needed.

I do however realize that professionals such as your good self need to follow the conventional path to get the job done and I am not really arguing with that point of view.

[Manfred M]

I get your drift in that only "the popular software" is worthy of mention here. It's a great pity that less popular (RawTherapee) software has ProPhoto as it's default working space and, beyond that, allows you to define and select any gamma you like in the output color space. For dabblers like myself, such a freedom to "influence the settings" is a delight, should it ever be needed.

I do however realize that professionals such as your good self need to follow the conventional path to get the job done and I am not really arguing with that point of view.

We all have different reasons and interests in photography. I'm trying to create compelling images and want enough technical knowledge to do that. Your interests seem to be far more on the technical side and are interested in pushing and prodding pixels to understand what happens.

As a more mainstream photographer, I focus my knowledge on what helps me create stronger images and use the generally accepted tools because it gives me a common base to work on with other like-minded photographers.

[billtils]

... Sometimes these can be changed (colour temperature cannot) ...

Manfred, did you mean this? Or have I not picked up the intent correctly? Changing colour temperature of a TIFF (or JPEG or any other format I've looked at) is as simple as adjusting the Kelvin slider in the editor.

[george013]

Manfred, did you mean this? Or have I not picked up the intent correctly? Changing colour temperature of a TIFF (or JPEG or any other format I've looked at) is as simple as adjusting the Kelvin slider in the editor.

I think it depends on the converter. In the Nikon software white balance and exposure compensation are disabled for non-nef. I think their procedures are based on the raw file.
In Photolab I can change them.

George

[Manfred M]

Manfred, did you mean this? Or have I not picked up the intent correctly? Changing colour temperature of a TIFF (or JPEG or any other format I've looked at) is as simple as adjusting the Kelvin slider in the editor.

No this is not a mistake and is in fact correct.

In the raw converter, no colour space has been assigned to the data so any colour related variable can be changed with no consequence in final image quality as the calculations are done on the base data. We have three sliders that affect the colour temperature / white balance that are labelled exposure, temperature and tint. If you understand your colour models, these are nothing more than the L, A and B channels of the LAB colour space. The "A" channel affects the colour temperature only. The beauty of this approach on raw data is that we can change the exposure without affecting the saturation of the colours.

Once we have assigned an RGB colour space the raw converter will have assigned a specific R, G and B value to each pixel based on whatever colour space has been assigned. The same type of operation occurs if we assign our working colour space as CMYK or LAB.

Sticking with one of the RGB colour spaces, changing the exposure recalculates each pixel value and affects the saturation as well as the exposure and to some extent the noise, Adobe still calls the Blue - Yellow adjustments "temperature", but assigns a numeric value. Adobe replaces 2000K to 50000K values in the raw convertor to a scale from -100 to +100. PhotoLab goes from the same 2000K to 50000K values when working with the raw data to "Cooler" and "Warmer" settings without assigning a numeric value we can see.

If you want to test this yourself, take a raw file and convert it to a 16-bit TIFF and see the effect of the "Temperature" slider when it works on raw versus image data. Minor tweaks don't show much of a difference, but push the sliders toward the extremes and the differences become quite obvious.

[xpatUSA]

No this is not a mistake and is in fact correct.

In the raw converter, no colour space has been assigned to the data so any colour related variable can be changed with no consequence in final image quality as the calculations are done on the base data. We have three sliders that affect the colour temperature / white balance that are labelled exposure, temperature and tint. If you understand your colour models, these are nothing more than the L, A and B channels of the LAB colour space. The "A" channel affects the colour temperature only. The beauty of this approach on raw data is that we can change the exposure without affecting the saturation of the colours.

Once we have assigned an RGB colour space the raw converter will have assigned a specific R, G and B value to each pixel based on whatever colour space has been assigned. The same type of operation occurs if we assign our working colour space as CMYK or LAB.

Sticking with one of the RGB colour spaces, changing the exposure recalculates each pixel value and affects the saturation as well as the exposure and to some extent the noise, Adobe still calls the Blue - Yellow adjustments "temperature", but assigns a numeric value. Adobe replaces 2000K to 50000K values in the raw convertor to a scale from -100 to +100. PhotoLab goes from the same 2000K to 50000K values when working with the raw data to "Cooler" and "Warmer" settings without assigning a numeric value we can see.

If you want to test this yourself, take a raw file and convert it to a 16-bit TIFF and see the effect of the "Temperature" slider when it works on raw versus image data. Minor tweaks don't show much of a difference, but push the sliders toward the extremes and the differences become quite obvious.

An interesting view relating CIELAB axes to color temperature and tint. Not having Adobe or PhotoLab, I am unable to test per Manfred's suggested method.

Nevertheless, I imagine that if I converted x,y values along the temperature line below, I would get a straight line parallel to the a* axis in CIELAB space, thereby confirming Manfred's statement.

[billtils]

No this is not a mistake and is in fact correct.

---

If you want to test this yourself, take a raw file and convert it to a 16-bit TIFF and see the effect of the "Temperature" slider when it works on raw versus image data. Minor tweaks don't show much of a difference, but push the sliders toward the extremes and the differences become quite obvious.

The expanded account shows what you mean, but anyone with the appropriate software can alter the Kelvin values for a TIFF or JPEG and see a change in the image, which is not in alignment with the original wording.

[dje]

I don't think there is a direct relationship between CIELab and Color Temperature adjustment. The DCP camera profile contained in a DNG file can include an LUT based on the CIELab color space and this is used for fine tuning of the color profile. The DCP profile also includes

One or two Color Matrices (two different illuminants) which are used for conversion between White Balance multipliers and Color temperature and Tint.

One or two Forward Matrices (two different illuminants) which are used for initial color calibration of the raw data.

This stuff is all documented in the DNG Specification document published by Adobe. They also provide SDK code for use by software developers who want to work with DNG files but that's way beyond my capabilities!

Coming back to Bill's question..

White Balance is fundamentally about applying multipliers to the R and B channels such that a uniform gray color has equal R G and B values for the illuminant under which the photo was taken. The As Shot values are included in the raw file metadata. The traditional user interface in editing software comprises two sliders - Color Temperature and Tint. The starting values for these are calculated from the starting values for the WB multipliers and when adjustments are made with the sliders, these are converted back to adjusted WB multiplier values to be applied to the R G and B channels.

Once a file has been converted to tiff or jpeg, the original WB multipliers and Color Temperature info are gone. So when you adjust the WB sliders for one of these files, it's a relative adjustment only as indicated in Manfred's earlier post (-100 to 0 to 100). The WB channel multipliers are still changed in a way consistent with changing values of Color Temp and Tint but there is no reference to actual equivalent Color Temp (Correlated Color Temperature to be precise).

Dave

[billtils]

The WB channel multipliers are still changed in a way consistent with changing values of Color Temp and Tint but there is no reference to actual equivalent Color Temp (Correlated Color Temperature to be precise).

Dave

Thanks Dave - that's what I was trying to say but didn't manage to do so as clearly as you have done.

[george013]

I just compared a nef and a jpg in Photolab. Looking at the whitebalance slider with a nef I get the colour temp and with a jpg just cooler and warmer.

George

[xpatUSA]

I don't think there is a direct relationship between CIELab and Color Temperature adjustment.

Dave

I tend to agree. Here is a CIELAB plot of four Color Temperatures 10000, 6000, 4000 & 3000K (white dircles):



You can enter color temperature directly into this calculator and get values for several color spaces (excluding Melissa ).

I left the white point at D50.. In LAB(D50), 5000K is very close to a*,b* = 0,0 ...