First, it is important to understand that the dng exported by PureRAW (or PhotoLab) is not a standard dng, but a linear dng with its own characteristics. It is not equivalent to a “standard” dng that would come directly from a camera whose raw files are of type .dng, or created via Adobe dng converter from a raw. These dngs are exactly the equivalent of a raw and are obviously not demosaiced.
As you have noted, the linear dngs of PureRAW (but this is not a format specific to DxO) are demosaiced files. That is to say that the luminance values recorded for each photosite of the sensor were used to determine the red, green and blue values corresponding to this photosite (according to the Bayer or X-Trans matrix). At the same time, the noise and lens corrections were taken into account. And these operations are done in the color space of the sensor.
As the value of each of the R, G, B layers of this photosite is defined, this explains why these dngs have an average size three times larger than a raw or a tiff. This is exactly what the dng produced by PureRAW is, which remains, I remind you, in the color space of the sensor, and therefore has no integrated profile: at this stage it is simply impossible.
To transform each of these photosites into a pixel, it will be necessary to subsequently combine the values of each of these layers by applying a white balance, a color rendering, etc. in the workspace of the software in which we will import this dng (Wide Gamut for DxO, Melissa for Adobe…).
And finally we will only be able to assign (integrate) an sRGB, Adobe, etc. profile when we export a tiff or jpeg.
Afterwards, it all depends on the software you use to read the metadata of linear dngs.
Some display the profile of the integrated thumbnail. Others (ACDsee for example) do not display a profile: N/A, exactly like for raw.
If PhotoLab displays sRGB in the metadata palette, it is because it corresponds to the setting made in the camera by the user for direct jpegs (jpeg or raw+jpeg). PureRAW will use this metadata for jpg and/or tiff exports. PhotoLab indicates sRGB for your dng files, for mine it indicates Adobe RGB because that is the choice I made in the settings of my cameras. And the profile integrated into the jpg/tiff files exported by PureRAW will be sRGB for you, Adobe RGB for me. Unlike PhotoLab where you can choose the integrated profile.
Have you tested this for yourself? The following is not true. It is not what happens. There is no documentation for PureRaw that describes this, either.
Of course I tested it myself!
By the way, you can also check if you have Lightroom or Photoshop:
Import a raw file directly into Adobe/ACR (preferably with saturated colors) and do the same with the dng from PureRAW. With the same settings of course. You will see absolutely no color difference between the two, whereas if the dng had an integrated profile (especially if it is an sRGB!), the difference should be clear. Obviously, this assumes that you are using a calibrated screen, covering Adobe RGB if possible.
And you can trust the statement that the linear dngs from PureRAW are in the color space of the sensor, and the jpeg/tiffs correspond to the camera setting: this information comes directly from DxO, following a question asked by me during a press conference, the user manual not being very precise on this point.
Note: for the characteristics of the linear dng, they have been public for a long time. And the user manual is clear since it clearly states: “Generates a Linear DNG file that retains the characteristics and reversibility of the original RAW file”, which would not be the case if a profile (especially sRGB) was integrated.
The DNG LinearRAW image data is in the camera native color space.
The ColorSpace tag found in EXIF data is just a hint written by the camera to indicate the prefered color space for the final target, as set by the user in camera settings.
I think this makes the whole discussion void, if not somewhat funny.
See also quoted old DxO page in Two export modes to DNG - TuTo DxO. The current version of the linked DxO page for some reason does not contain this information anymore, perhaps because they thought it was too obvious
Note that DNG specs does not actually define LinearRAW but it uses the term (btw, quite typical for Adobe specs).
First, I have never used sRGB as the preferred color space in camera. It’s one of the things I changed on day one and never had a reason to change again. So this label is not coming from the camera. Second, this label is not coming from the system profile or the display profile. For what it’s worth, PL8 does the same thing when exporting as DNG.
Is all of the information in the original RAW file in the linear DNG? Maybe it is and maybe it isn’t. It isn’t by definition. DxO says it is, so it must be so, because they never misspeak, mislead or make errors? If it’s being mapped to sRGB in the interim, everything outside of sRGB could get blended in using perceptual or relative modes or simply clipped completely. But let’s just assume everything in the linear DNG image is a-okay. It is still either a bug or just not good practice (for obvious reasons looking at this thread) to assign a color space to what is effectively a RAW image in the metadata. Load the same file into other image editing software and it is assigned RAW status and does not have a color space defined.
The ColorSpace tag (0xa001) displayed by PL in ‘METADATA Exif’ panel for the DNG file generated by PR probably refers to JPEG preview image embedded in DNG container along with LinearRAW data. You may safely forget it, although I agree it may be confusing. Let me repeat, the LinearRAW data is in the native camera color space, 3x16-bits per pixel, so there’s nothing to worry. In PL, exporting to linear DNG does not let you set the output color space exactly for the same reason – it is predefined to be “camera native”.
1 Like
Stenis
(Sten-Åke Sändh (Sony, Win 11, PL 6, CO 16, PM Plus 6, XnView))
47
I don´t use Apple-computers or monitors at all and that goes for all Apple-products including phones. How come you are presuming something like that?? You are free to use Display P3 on any hardware-calibrated monitor and software that supports it.
Apple computers have had their nisches in photo and music (MIDI) but industry standard is still Windows and I use Windows 11 and a Display P3 -profiled Benq-monitor and I have already described my reasons to do that. All my image
-files are postprocessed with a Display P3 bias and exported with Display P3 ICC for screens. Every file I print is developed with Display P3-profile and it works perfectly fine. I just happen not to like the Adobe RGB-color space and when I print my pictures on my Epson P900 I am free to print in what ever color space I prefer. I would never use Adobe RGB if I wasn´t going to outsource a print of some reason - period.
I think you will have to open your eyes to be able to see the present color space world that is rapidly changing. Display P3 can be used by anyone today and is not something locked exclusively to the Apple-world. The days are gone when sRGB was the only standard for both displays and amateur printing and even printing can be done with Display P3 if that happens to be your color space of choice and you control the printing processes by yourselves. There is nothing right or wrong here - just a choice to do or not to do, after your likings. Why is that so hard to take in the year 2025?
Never mind. I’m good, and everything is cool . There is no perceptible difference. Thank you for the conversation; I’m not in the habit of just believing what internet people tell me. DxO would do well to remove the color space reference from the metadata palette for all DNG images. All it does is make people mad.
The following is what happens when my in-camera JPG setting are set to sRGB or Adobe RGB. The RAW file Color Space entry is sRGB or Uncalibrated, respectively. In PL8, with the No Corrections Preset and Denoise & Optical Corrections only for linear DNG export, the Color Space entry is sRGB or Uncalibrated, respectively. That is, PL8 does NOT change the Color Space entry. Further, as expected, no ICC color profiles are ever embedded in the exported linear DNG files.
Note that most “image files” usually contain image data for several images, like RAW data, linear raw data, jpeg preview and thumbnail versions, etc. The ColorSpace tag just says what was used for the output color space to generate some “final” image, the embedded preview jpeg in DxO linear DNG case. Both RAW and LinearRAW are in camera native space and the ColorSpace tag is irrelevant for them. They are “intermediate image files”, not the “final” ones.
A fitting post-mortem for an unfortunate thread. Could add “not a bug” or “useful after all” I suppose. Kudos to you, @gerarto, @thomas001le, and others for calmly dispatching this.
. There is no perceptible difference. Thank you for the conversation; I’m not in the habit of just believing what internet people tell me. DxO would do well to remove the color space reference from the metadata palette for all DNG images. All it does is make people mad.