Thank you for sharing the link. That’s probably the white paper mentioned in the link provided by @Egregius above.
Dynamic Range Compression (DRC) – similar curves are used by camera makers for lossy RAW compression. For Nikon see https://www.photonstophotos.net/NikonInfo/NEF_Compression.htm.
In a sample HiFi DNG PL9.6 output, the Linearization Table has 1024 quantization levels. It starts with 0 2 5 7 9 11 14 16 18 20 23 and ends with 64514 64641 64769 64896 65024 65151 65279 65407 65535.
To quote the DxO WP, “We apply JPEG XL with a near-lossless quality setting”. In fact, they use JXLDistance = 0.01, minimal non-zero, as mentioned above. Normally it wouldn’t compress that good but the DRC makes the trick.
I look at it this way. If compressed DNGs had been available in PhotoLab 3 with PRIME noise reduction, and I had converted all my photos then, how would I feel now with DeepPRIME 3 or XD3 available?
DeepPRIME variants, plus Lens Sharpness Optimisation work only with original RAWs (or direct DNG conversions of same). You’re throwing away any future technology that DxO may produce that relies on the original RAW data.
If “good enough” is truly OK for you, then go for it, but I cannot see myself throwing away the originals. Perhaps, I guess, the originals could be archived onto offline storage, but then comes the other problem…
Are you going to bulk convert images? Which settings are you going to use? When I process photos in PhotoLab, I do start with and, in most cases, keep my standard settings. But… I also tweak them for some images that benefit. If you do a bulk conversion you are also throwing away this ability.
1 Like
On removing non-technical corrections in DNG output.
The DxO FAQ article explains it – see the link provided by @Egregius above.
The “non-technical” corrections are done in DxO working colorspace (DxO Wide Gamut by default, non-linear), so translating them back into camera native colorspace can easily produce unwanted color shifts. Probably too many people complained about it and DxO decided to use in Linear DNG output only “technical corrections” (i.e. those done in native colorspace, listed in FAQ and 9.6 release notes).
1 Like
Yes I’m aware of the limitations, and for the photography I’m doing as a hobbyist and the kind of things I like visually, DxO’s current algorithms are good enough and I can live with having them baked in the resulting linear DNGs.
The UX of bulk converting might be a blocker, I’ll have to see how it goes. I think the main issue might be for images that I’ve edited with PhotoLab and which have a .dop already. I’ll need to try it.
Um, OK. So, what are the practical implications?
I’m an event pro. 99.99% of my images are disposable after I deliver finished JPEGs to my clients. I keep them in case a client comes back to me with a request, or in case I might want to use them on my website. For these purposes, even the JPEGs would suffice, but, being a perfectionist, I want to have the editable color depth and dynamic range of the original RAWs, and lossy compressed DNGs provide this in a much smaller file. BUT, I want to have ALL the adjustments I made in PhotoLab included.
2 Likes
I don’t want to sound flippant, but could you not find some other software that turns, say, 16-bit TIFFs into compressed DNGs? Both are industry standards. As has been said before, DxO didn’t invent anything here. They did, perhaps, tune some parameters to feed into the compression engine. You could come up with your own parameters to better match what you want.
As it is, you’re using a function for a purpose other than what it was intended for.
I don’t want to sound flippant, but can you not imagine why it would be helpful to have this function built into PL rather than having to export 16-bit TIFFs, import them into Lightroom Classic, export compressed DNGs from there, import those into Lightroom Classic, then delete the TIFFs?
And, doncha just love it when someone tells you you shouldn’t want what you want, without actually understanding why you want it?
2 Likes
Speaking only for myself, I’ve yet to find a use for linear DNG, as it seems to function as an intermediate stage between RAW and a final RGB product. With that in mind, I think I understand why DxO changed its export options for DNG. Still, this discussion is convincing me that I have more to learn about DNG and shouldn’t rely on assumptions. Those who use DNG in their workflow need to be listened to. This discussion is also convincing me that DxO should be giving us more options for export formats, not fewer. I suppose part of the problem is that formats like JPEG XL and HEIC/HEIF are not as universally supported as TIFF and DNG presently are - but someone has to lead the way, and DxO should at least be following others’ lead if they can.
The current linear DNG export option is clearly a feature and not a bug. You (@Jacques4242) and others have made a good case for additional export options. Why not launch a new formal feature request? I would vote for it. The current (minimally) lossy linear DNG, but with all corrections applied, would be a good start.
What about adding lossless linear DNG as well, using standard JPEG XL compression, without DxO’s dynamic range compression? Might appeal to the archival, reproduction, and forensic crowd, and would stay clear of any proprietary issues.
Down the road I would also like to see export options for the JPEG XL format, both lossy and lossless, but that’s another story.
1 Like
Based on my observations of how DxO tend to work on the ‘our way or the highway’ principle I’d suggest that these options are far more likely to happen than for DxO to restore the previous options for DNG export.
Almost as much as someone who self identifies as a perfectionist and uses phrases like “WTF” and oodles of ? and ! marks.
You didn’t so much say you wanted it as almost demand it and call DxO’s parentage into question for daring to remove a function you were using in ways they never intended.
At least I offered a suggestion. And with that, I’m out. This thread is being muted.
1 Like