Ad
Ad
Ad
Pages: [1] 2 3 »   Bottom of Page
Print
Author Topic: Lightroom-DxO Round-Trip  (Read 3604 times)
Chris Kern
Full Member
***
Offline Offline

Posts: 126


« on: May 23, 2014, 06:00:11 PM »
ReplyReply

I've been playing around with DxO Optics Pro 9.5's new trick: the ability to import a raw file from Lightroom and return a linear DNG.  If I understand this properly (as usual, the explanatory material on the DxO website is less than pellucid), by emitting a DNG, DxO Optics Pro allows a Lightroom user to take advantage of DxO's sophisticated optical testing and correction technology while still retaining all the functionality of LR—except, of course, Adobe's demosaicing algorithms.  For the moment, at least, I'm only interested in DxO's automated sensor-and-lens corrections, with all additional processing performed in LR.

I'm encouraged by the initial results.  I've attached a small center crop from a much larger image made with a Nikon D800E and Nikon's 24-120mm f/4 zoom.  The first attachment is a 1:1 JPEG of the raw file as demosaiced by LR after applying LR's automatic lens corrections.  The second attachment is the same raw file demosaiced by DxO after applying DxO's automatic corrections for the sensor and lens.  I didn't do any manual sharpening or make any other corrections with either tool since my objective is to compare how the two perform the raw conversion and correct for known optical imperfections in the capture.  I've only worked on a few images so far, and only those shot with this particular sensor-lens combination, so it's too soon to come to any definitive judgment.  But in every test I've made, the DxO image looked better than the LR equivalent—sometimes dramatically so.

My only concern is that I may be giving up some of Lightroom's functionality by having DxO perform the conversion and turn over the linear DNG to LR.  It appears LR can still properly modify the white balance of the DxO-emitted DNG.  Is there anything else I might be giving up, or should otherwise be concerned about?
« Last Edit: May 23, 2014, 07:39:30 PM by Chris Kern » Logged
Steve House
Full Member
***
Offline Offline

Posts: 226


« Reply #1 on: May 23, 2014, 06:25:55 PM »
ReplyReply

I too have been curious about this since 9.5 was announced.  Consider the workflow:  Shoot RAW (NEF).  Import into LR, converting to DNG on import. Export DNG to DxO and process.  Re-import DNG from DxO back to LR.  Export from LR to image file (JPG, TIFF, etc) or print directly from LR.  What engine is actually doing the demosaicing, Adobe's, DxO's, an additive combination of both?  I've always thought the DNG was itself a RAW file format, with the colour conversions only getting 'baked in' when a true image file was created from it or it was printed.  That would seem to suggest such a workflow would be using DxO's optical corrections but Adobe's colour processing.  Anyone really know???
« Last Edit: May 23, 2014, 06:28:26 PM by Steve House » Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9191



WWW
« Reply #2 on: May 23, 2014, 06:28:51 PM »
ReplyReply

In a nutshell, the Linear DNG isn't raw (it's far less raw). It's rendered but in a linear space. Could have been a TIFF too, don't let the DNG part out give you the impression it's the same as DNG in from Lightroom.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Steve House
Full Member
***
Offline Offline

Posts: 226


« Reply #3 on: May 23, 2014, 06:37:50 PM »
ReplyReply

So what happens if you do additional processing in LR after bringing it back from DxO (compared to never sending the DNG to DxO in the first place)?
Logged
Chris Kern
Full Member
***
Offline Offline

Posts: 126


« Reply #4 on: May 23, 2014, 06:39:22 PM »
ReplyReply

In a nutshell, the Linear DNG isn't raw (it's far less raw). It's rendered but in a linear space. Could have been a TIFF too, don't let the DNG part out give you the impression it's the same as DNG in from Lightroom.

Understood, Andrew.  I'm aware the file is "far less raw."  But my understanding has been that, for example, a TIFF has the color balance baked-in, but that a linear DNG can still be color-corrected.  (And, indeed, I can't detect any difference when I fiddle with the color sliders in LR.)  Am I missing something here?  And even if I'm not, is there something else, perhaps unrelated to color, that I'm missing?  I don't understand enough about the different semantics of the TIFF and DNG file formats to know how to perform definitive controlled tests.
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9191



WWW
« Reply #5 on: May 24, 2014, 10:42:36 AM »
ReplyReply

I'd think of the linear DNG just as I would a linear TIFF as AFAIK, the data would be the same (high bit, linear). Different format container, still baked and rendered, less so of course than a JPEG from the camera or a TIFF processed by you in any other raw converter. I don't see how the baked linear DNG would provide any major capabilities over the same data in a TIFF.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Chris Kern
Full Member
***
Offline Offline

Posts: 126


« Reply #6 on: May 24, 2014, 11:21:37 AM »
ReplyReply

I'd think of the linear DNG just as I would a linear TIFF as AFAIK, the data would be the same (high bit, linear). Different format container, still baked and rendered, less so of course than a JPEG from the camera or a TIFF processed by you in any other raw converter. I don't see how the baked linear DNG would provide any major capabilities over the same data in a TIFF.

Sigh.  I'm still struggling to understand this—and why, other than marketing hype, DxO is making such a big deal about the ability of the latest rev of Optics Pro to pass a linear DNG back to Lightroom.  (Of course, it may indeed just be marketing hype.)

I just did a web search, and found a couple of explanations by Eric Chan which I think may be relevant to this discussion.

From a 2011 discussion:

Quote
DNG is indeed very much like TIFF.  It’s actually a set of TIFF extensions, with tags to describe things like white balance, color profiles, and various calibration data needed for processing raw images (as opposed to already-rendered images like regular jpegs and tiffs). . . .

At its most essential level, DNG is just a container of image data and its associated metadata.  It can hold raw image data (a.k.a., scene-referred data), and it can also hold rendered image data (a.k.a., output-referred data).  For raw data, the data can be in the mosaic form, or it can be demosaiced (so-called “linear DNG”).

(Source: http://www.natcoalson.com/blog/2011/11/29/my-adobe-dng-chat-with-eric-chan/)

And from a 2008 post by Eric on the DPReview website:

Quote
A usual TIFF file that comes out the back end of a raw converter has already been rendered, i.e., it has been mapped to a standard color space, it has been tone mapped, white balancing has been done, etc. More technically, the image is output-referred.

In contrast, the linear DNG is still scene-referred and can still benefit from many of the operations typically performed by a raw converter, such as white balance, the application of a camera color profile, HDR compositing, etc.

So the underlying internal file format looks similar, but the actual image contents and the types of operations that can be applied to that image are quite different.

(Source: http://www.dpreview.com/forums/post/30148341)

Now I think what these mean is that while both TIFF and DNG are just containers, one thing that distinguishes a linear DNG from a TIFF is that it contains additional metadata which allow some processing that is usually only performed during the raw conversion process.  Does that sound right?  If so, it might account for the claims DxO is making—and give them some credence.
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9191



WWW
« Reply #7 on: May 24, 2014, 12:10:02 PM »
ReplyReply

Metadata but more so, less processing hence the idea (and that's up to debate) that this is scene referred not output referred data. See:
http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf
It's rendered so not raw, but minimally processed. It's certainly more scene referred than output referred. Some operations may be easier to conduct, white balance comes to mind. But it isn't raw and that's the big deal IMHO. Don't let the DNG part of the name give you an idea it's a digital negative as we think of a raw neg. You've baked those pixels but to a lesser degree than using other tools on the data.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
john beardsworth
Sr. Member
****
Offline Offline

Posts: 2833



WWW
« Reply #8 on: May 24, 2014, 12:28:23 PM »
ReplyReply

Anyone think there's a good reason why they don't return a proper DNG with their processing instructions added as metadata?
Logged

digitaldog
Sr. Member
****
Offline Offline

Posts: 9191



WWW
« Reply #9 on: May 24, 2014, 12:29:56 PM »
ReplyReply

Anyone think there's a good reason why they don't return a proper DNG with their processing instructions added as metadata?
Outside their processing, how would it be used? But if they are supposed to do this, they most certainly should.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Chris Kern
Full Member
***
Offline Offline

Posts: 126


« Reply #10 on: May 24, 2014, 12:45:52 PM »
ReplyReply

It's rendered so not raw, but minimally processed. It's certainly more scene referred than output referred. Some operations may be easier to conduct, white balance comes to mind. But it isn't raw and that's the big deal IMHO. Don't let the DNG part of the name give you an idea it's a digital negative as we think of a raw neg. You've baked those pixels but to a lesser degree than using other tools on the data.

Yup.  I understood that going in.  But to bring this back to the starting point, I've been favorably impressed for years with the results of DxO Optics Pro's automated optical corrections.  Lightroom's lens correction module works pretty well, but I've found DxO to be generally superior—at least with the sensor-lens combinations I've tried.

On the other hand, I prefer LR (with an occasional but increasingly rare visit to Photoshop) for other post-processing.  So I turn off all adjustments in DxO except the corrections directly related to the company's sensor-lens database, and use LR for everything else.

Until now, that has meant rendering in DxO Optics Pro and importing a TIFF into Lightroom.  Aside from the awkwardness of the extra manual step, there were things that definitely didn't work right when I worked on the DxO TIFF in Lightroom.  For example, white balance adjustments simply didn't match what I could do in LR with the raw file.

It seems (I'm not confident in the comprehensiveness of my initial judgment) that DxO's new ability to emit a DNG has solved both problems.  I can hand off the raw to Optics Pro directly from LR (DxO provides a new LR plug-in for that) and the Lightroom sliders seem to work pretty much the same way on the DxO-emitted DNG as they do on the raw file, modulo the inherent difference in the way the two programs perform the raw conversion.  While I may no longer be working on the raw file, it's still there if I'm not happy with the DxO result.  But mostly, the optical corrections made by DxO as part of its rendering process seem like a win to me.

Now the ideal arrangement from my perspective would be for Adobe to license DxO's technology and use it to make sensor-lens corrections to the raw file on-the-fly.  But short of that, half a half-backed loaf is better than a fully-baked loaf.  If that makes any sense.
Logged
john beardsworth
Sr. Member
****
Offline Offline

Posts: 2833



WWW
« Reply #11 on: May 24, 2014, 01:09:37 PM »
ReplyReply

Outside their processing, how would it be used? But if they are supposed to do this, they most certainly should.

It would be for their processing, just like ACR adjustments. I'm surprised they didn't follow that route.
Logged

Chris Kern
Full Member
***
Offline Offline

Posts: 126


« Reply #12 on: May 24, 2014, 02:36:28 PM »
ReplyReply

It would be for their processing, just like ACR adjustments. I'm surprised they didn't follow that route.

As far as I can determine, DxO Optics Pro operates directly on the sensor data from the raw file and makes adjustments on-the-fly, similarly to Lightroom (i.e., "non-destructive parametric editing"), then produces a rendered file only if the user wants it to emit one.  The major change from previous revisions of the product is that v. 9.5 now can produce linear DNG files as well as TIFFs and JPEGs.  Plus, they've added some functionality and a Lightroom plug-in to allow the user to (1) tell DxO which raw file(s) to operate on from within Lightroom (I'm not sure whether that offers any advantage over Adobe's exiting "Edit-In" facility) and (2) pass the DxO output file back to Lightroom.
Logged
Misirlou
Sr. Member
****
Offline Offline

Posts: 657


WWW
« Reply #13 on: May 24, 2014, 03:42:10 PM »
ReplyReply

It's just more convenient to work with this new DxO path, at least for me.

You might also consider noise reduction in DxO. Depending how noisy your image is, and what you intend to do with it, the DxO "prime" reduction process is pretty amazing. It takes a very long time to process, but I haven't seen better results with any other system. Useless for a wedding photographer, but if you've got a single image with a lot of noise, or a small group of them, it's worth the hassle.

I recently shot a set of HDR sequences for panorama stitching. After much experimentation, I determined that I got the most out of them by converting the raws, with lens correcting and denoising in DxO. Then I tried several flavors of HDR, with the NIK process (also launched from Lightroom) coming out the best with that particular set of photos.

I think a lot of the complaints about DxO white balance come down to their default camera profiles. You get an easier profile selection (or custom creation) with Adobe, and their recent default camera profiles are quite good (with my cameras at least). If you're willing to come up with better user profiles for DxO, you could probably do just about as well.

But no matter how good DxO gets, LR's DAM will always be important to me, as well as easier creative controls with LR's tonal adjustments.
Logged
Steve House
Full Member
***
Offline Offline

Posts: 226


« Reply #14 on: May 24, 2014, 03:42:53 PM »
ReplyReply

When I import a NEF camera raw file into Lightroom with LR converting it to DNG on import, is the resulting file a "linear DNG?" What other types of DNG are there?
Logged
dennbel
Newbie
*
Offline Offline

Posts: 16


« Reply #15 on: May 24, 2014, 04:05:16 PM »
ReplyReply

I've often wondered just how accurate the DXO correction is for MY lens and My sensor. How many do they test and do they average them? Same model lens and sensor from batch to batch have variances. So what they get, may not be entirely accurate for my lens and sensor. Just thinking out loud.
Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5503


WWW
« Reply #16 on: May 24, 2014, 07:44:28 PM »
ReplyReply

Anyone think there's a good reason why they don't return a proper DNG with their processing instructions added as metadata?

Because DxO's image corrections need to alter pixels with their algorithms for lens corrections. ACR/LR could use DxO's algorithms if stored in metadata unless DxO gave (licensed) the algorithms to Adobe...so, DxO must render the raw pixels into RGB pixels to run the distortion/lens corrections on.

Linear DNG for all practical purposes is a semi-raw file format. The main difference is dealing with the camera color to processing profile conversions–with Linear DNG, you can't go upstream from that initial color transform.

That and the fact that a Linear DNG limits the full capability of PV 2012's highlight preservation...an L/DNG is really close but not exactly the same as raw.

I seriously doubt that Adobe (read Thomas) will open the ACR/LR processing pipeline to the extent that 3rd parties could plug into the pipeline...so, I suspect it'll always require a new file image be spawned off either as a rendered TIFF or L/DNG. The original raw will still be preserved and treated as read only.
Logged
Charlene McKinnon
Newbie
*
Offline Offline

Posts: 6


« Reply #17 on: May 24, 2014, 11:38:10 PM »
ReplyReply

If you want to take advantage of all the features of DxO which normally require a raw file, except use Lightroom for rendering the raw image data, you can:

  • export a tiff from Lightroom
  • use exiftool to transfer raw metadata to tiff file
  • open the tiff file with DxO
  • make lens corrections and/or any others you want to - as you would to a raw file
  • export a jpeg (or linear DNG if you prefer...)
  • Import the jpeg (or DNG, or TIFF, or ...) in Lightroom

This will work perfectly as long as you don't bake distortion/lens corrections into the tiff beforehand, or crop, or make other corrections which are best reserved for DxO.

Obviously if you will be doing this a lot, a batch file or plugin or some way to automate repetitive steps would be invaluable.
Logged
Ligament
Full Member
***
Offline Offline

Posts: 186


« Reply #18 on: May 25, 2014, 03:29:10 AM »
ReplyReply

Does DXO Optics Pro 9.5 *still* use Adobe RGB as its internal working colorspace? If so, you could loose a lot of valuable data exporting from lightroom (profoto working space) to DXO (adobe RGB) back to lightroom (now with an adobe RGB file).

This has always been a peeve of mine with DXO. Why on earth would they use Adobe RGB as the internal working space? Why throw all that data away?
Logged
Damon Lynch
Full Member
***
Offline Offline

Posts: 135


WWW
« Reply #19 on: May 25, 2014, 03:49:58 AM »
ReplyReply

Thanks everybody for explaining linear DNG.

If you want to use LR but like DxO's optical corrections, there is also DxO Viewpoint.  It's so much cheaper. It is a real time-saver when it comes to correcting the volume deformation caused by wide angle lenses, particularly useful for photos with people in them.

I remain bemused by the apparent fact that in this multi-billion dollar industry, one set of people (a company's own optical engineers) know perfectly well the optimal algorithms to improve image quality in post-processing, yet the set of most important people in post-processing (Adobe, DxO, et al.) must themselves develop algorithms that only approximate the ideal corrections. Would companies like Canon and Leica suffer if these optical correction algorithms were mandated to be open access?
Logged
Pages: [1] 2 3 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad