Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: David Eckels on May 26, 2013, 02:59:50 pm
-
I had heard that Nikon's software used to convert RAW images can extract proprietary information embedded in the NEF raw file. I must confess I was a bit skeptical. My workflow, up until now, has been to shoot in 14-bit raw (NEF) and convert to the DNG format using Adobe's software. The NEF files were kept as a backup and all post-processing was done on DNG raw files. I decided to do a little experiment to see what kind of extra information might be present in NEF that could not be teased out of a DNG file or its equivalent when Adobe did the raw conversion. For each shot, I started with the original raw NEF and opened it in Lightroom 4.4. LR converts the NEF on the fly and it is the same whether using Adobe Camera Raw or the Adobe DNG Converter software as far as I can tell. Alternatively, I opened the NEF in Nikon ViewNX 2 and converted to 16-bit TIFF format before opening in LR4.4, thus any proprietary information would have the advantage of the Nikon raw conversion. For each comparison, I made comparable 100% crops. My first experience was with a file from the D7000, but I thought maybe more recent cameras would not show the difference, so I tried the same experiment with my D7100 and D800E cameras (shown here).
With the D7100, the shot was taken under fluorescent lighting following a flood in my laboratory (f/8, 1/100 sec, ISO 1600, Auto WB). Like with the D7000, there are definite differences in white balance, sharpness and noise between the two, with the Nikon ViewNX 2 providing a superior conversion in my opinion. See if you don't agree. I was amazed, frankly.
With the D800E shot of a Sedona sunset (f/8, 1/200, ISO 200, Auto WB), I thought maybe this phenomenon would be less intrusive using a higher end, higher resolution camera. Not so! The Nikon conversion again produced significantly better sharpness and the chromatic aberration was much less apparent. Note the odd red halo bordering the top of the mountain silhouettes along with noise and WB differences in the Adobe LR conversion. Canon software may behave similarly I understand.
Take home message: Raw may not be raw and the results depend on the converter used and this has caused me to change my workflow. Now, I convert the raw NEF to TIFF via ViewNX 2 and open the 16-bit TIFF for post-processing.
Do these results surprise anyone besides me? I know one could probably get a DNG file to look like the extracted NEF, but if raw is raw, why don't they convert the same? Maybe a trivial point, but interesting to me nonetheless.
-
Raw may not be raw and the results depend on the converter used and this has caused me to change my workflow. Now, I convert the raw NEF to TIFF via ViewNX 2 and open the 16-bit TIFF for post-processing.
Do these results surprise anyone besides me? I know one could probably get a DNG file to look like the extracted NEF, but if raw is raw, why don't they convert the same? Maybe a trivial point, but interesting to me nonetheless.
Hi David,
Raw is Raw, but every manufacturer knows its Raws best. In particular Nikon's software opens its NEFs to look pretty close to what the image looked like out-of-camera. If you use some of Nikon's advanced functionality (like Picture Controls and/or ADL) and you like how the images look OOC, then using V/CNX2 is a great time saver, because it gives you the OOC image as the PP starting point. I do. So I use Capture NX2 (I find VNX2 too limiting) to open and edit all of my NEFs. For 90% of them that's all I need. The remaining 10% get 'Opened With' PSCS for pixel or plug-in editing.
Cheers Jack
PS The images in your post do not show up in my browser.
-
Do these results surprise anyone besides me? I know one could probably get a DNG file to look like the extracted NEF, but if raw is raw, why don't they convert the same? Maybe a trivial point, but interesting to me nonetheless.
Raw is raw and must be demosaiced and converted into pixels with base level processing and conversion. Nikon uses their own SDK for processing, ACR/LR uses their processing pipeline. So, the differences should NOT be a surprise...what would be a surprise is if they actually matched.
Also, unless you optimized the image's sharpening and tone/color corrections (and others such as CA correction), then you haven't done your your work in ACR/LR...Nikon on the other hand does do default corrections that may make the image look "better" at default. All that means is Nikon is forcing their "look" on to your images while ACR/LR expect you to make the images look the way YOU want them to look.
As far as the NEF files containing "proprietary information embedded in the NEF raw file", yes...they do. Which is the problem with undocumented, proprietary raw files. Is that data important? Maybe, maybe not...since most embedded metadata unique to NEF files isn't used by ACR/LR, that data isn't important for processing in ACR/LR. ACR/LR does extract some metadata such as white balance (but uses it's own WB rendering) and things such as EXIF metadata for normalizing sharpening and noise reduction. It also extracts the lens data for use in ACR/LR's lens corrections–if you turn it on (it's off by default).
Seems you need to learn a bit more about raw image processing...in particular, how to use the ACR/LR controls for optimizing your images in ACR/LR.
-
Raw is Raw, but every manufacturer knows its Raws best.
Actually, that's a myth. The camera maker's software may be the best at matching in-camera previews and JPEGs and using the built in camera adjustments, but the guys working on ACR/LR camera support know the raw files pretty well too. In fact they have the experience of decoding upwards of 300 different camera sensors which kinda makes them experts...
The camera maker's software is clearly better at producing a match to the in-camera "looks" and processing at defaults. ACR/LR normalize all raw files to product the ACR/LR "defaults" which will not match the camera maker's processors...proper use of the ACR/LR processing controls can certainly be used to get fully optimized and rendered images...
-
...Do these results surprise anyone besides me? I know one could probably get a DNG file to look like the extracted NEF, but if raw is raw, why don't they convert the same? Maybe a trivial point, but interesting to me nonetheless.
No.
And the boys have already very eloquently explained why.
Tony Jay
-
Raw is raw and must be demosaiced and converted into pixels with base level processing and conversion. Nikon uses their own SDK for processing, ACR/LR uses their processing pipeline. So, the differences should NOT be a surprise...what would be a surprise is if they actually matched.
Also, unless you optimized the image's sharpening and tone/color corrections (and others such as CA correction), then you haven't done your your work in ACR/LR...Nikon on the other hand does do default corrections that may make the image look "better" at default. All that means is Nikon is forcing their "look" on to your images while ACR/LR expect you to make the images look the way YOU want them to look.
As far as the NEF files containing "proprietary information embedded in the NEF raw file", yes...they do. Which is the problem with undocumented, proprietary raw files. Is that data important? Maybe, maybe not...since most embedded metadata unique to NEF files isn't used by ACR/LR, that data isn't important for processing in ACR/LR. ACR/LR does extract some metadata such as white balance (but uses it's own WB rendering) and things such as EXIF metadata for normalizing sharpening and noise reduction. It also extracts the lens data for use in ACR/LR's lens corrections–if you turn it on (it's off by default).
Seems you need to learn a bit more about raw image processing...in particular, how to use the ACR/LR controls for optimizing your images in ACR/LR.
Jeff, thanks for the reply. I have read your book The Digital Negative and found it terrific. I understand your points and presented the above images unmodified, with LR optimizations purposefully turned off. But you are also saying, I think, that Nikon is applying noise, sharpening, CA, and WB adjustments to the same raw data as Adobe. Does this also apply to ADL? So, to put it another way, if I optimize in LR/ACR, I will not be losing any data that could only be retrieved by the Nikon software? This would be terrific as the dng format takes less space than tiff. I appreciate all the replies.
-
But you are also saying, I think, that Nikon is applying noise, sharpening, CA, and WB adjustments to the same raw data as Adobe.
Correct, unless you actually turn that stuff off.
So, to put it another way, if I optimize in LR/ACR, I will not be losing any data that could only be retrieved by the Nikon software?
You don't lose it, you just don't use it in ACR/LR. ACR/LR's processing doesn't use them because the pipeline don't use the Nikon SDK.
-
don't lose it, you just don't use it in ACR/LR. ACR/LR's processing doesn't use them because the pipeline don't use the Nikon SDK.
So the interpretation of "loss" is really Nikon's interpretation relative to a "blank slate" of Adobe's version? Jeff, thanks for this; helps us "youngsters" understand and ultimately makes us better photographers. Appreciate your teaching!You don't lose it, you just don't use it in ACR/LR. ACR/LR's processing doesn't use them because the pipeline don't use the Nikon SDK.
You don't use it, unless you turn on whatever adjustments (CA, noise, sharpening, etc) and THEN you CAN access it. Yes?
-
You don't use it, unless you turn on whatever adjustments (CA, noise, sharpening, etc) and THEN you CAN access it. Yes?
Confusing...not sure what you mean. ACR/LR only uses a small subset of embedded metadata embedded by a camera. ACR/LR uses it's own processing for all the things the camera maker's software can also do. But there's nothing proprietary that ACR/LR uses from the camera maker's SDK other than perhaps white balance (and even then, ACR/LR has it's own approach).
AS far as the secret about raw files, the secret is, there's really no secrets. Pretty much everything can be decoded. ACR/LR chooses not to use most of the proprietary metadata because it doesn't help the ACR/LR processing pipeline.
Also, with the exception of certain raw files (such as Sigma and early .CRW from Canon) when you convert to DNG, everything in the raw file is moved and retained into the DNG even if the DNG Converter doesn't use the metadata. It's quite possible that a DNG could be converted back to a valid raw file (I know this for a fact) but there u=is currently no reverse conversion options. However, converting a raw file to DNG does draw a line in the sand regarding the camera maker's software...once you convert to DNG, most non-Adobe raw processors can't process those files. Capture
one can, but only for those raw files that it can natively process anyway. Nikon and Canon software don't speak DNG at all (not surprising).
Bottom line, if you want to use ACR/LR for raw processing, don't worry about what the camera makers embed in the files...ACR/LR can get to what it needs to for processing–once a camera is supported.
-
You can convert your raw files to DNG but there is currently no reverse coding available.
Doesn't this show something about Adobe's way of thinking?
I have been converting all my raws to DNG but now regret doing so because this precludes me from using DPP.
I should have kept a copy of the original .cr2.
If a reverse coding thing was possible I'd certainly do it.
-
Confusing...not sure what you mean.
I guess what I am trying to say is that I was fooled into thinking (naively) that the Nikon conversion was "better" than the Adobe because, in effect, a Nikon "preset" was applied to the data, which involved Nikon's version of noise, sharpening, CA, and WB optimization. I could make up a similar set of presets in LR/ACR and thus I shouldn't have lost anything from the original raw data. Correct? Hope this is clearer.
One last question, I promise ;) Does this also apply to Active D-Lighting or is there some Nikon "magic" with respect to decoding ADL? I know you might say, raw is raw, and I can appreciate that! Thanks for hanging with me on this.
-
I guess what I am trying to say is that I was fooled into thinking (naively) that the Nikon conversion was "better" than the Adobe because, in effect, a Nikon "preset" was applied to the data, which involved Nikon's version of noise, sharpening, CA, and WB optimization. I could make up a similar set of presets in LR/ACR and thus I shouldn't have lost anything from the original raw data. Correct?
Correct...
Does this also apply to Active D-Lighting or is there some Nikon "magic" with respect to decoding ADL? I know you might say, raw is raw, and I can appreciate that! Thanks for hanging with me on this.
ACR/LR doesn't use Active D-Lighting settings and ignores them. Not being a Nikon shooter, I really don't know what Nikon is doing but ADL seems useful when shooting JPEGs to tame the contrast range of the scene...but I'm pretty sure the raw file is just the raw file and the post processing (either in-camera or in software) is where the processing happens. But I could be wrong...
-
ACR/LR doesn't use Active D-Lighting settings and ignores them. Not being a Nikon shooter, I really don't know what Nikon is doing but ADL seems useful when shooting JPEGs to tame the contrast range of the scene...but I'm pretty sure the raw file is just the raw file and the post processing (either in-camera or in software) is where the processing happens. But I could be wrong...
Jeff, I believe ADL is similar to Canons Highlight Tone Priority...both intended to protect highlights. That is, for jpegs, it will underexpose the scene to protect the highlights, then apply post processing adjustments to bring them up without clipping, I believe it is also possible they may apply a different curve.
The RAW file is the raw file. However, if ADL/HTP are turned on, the exposure will be effected....which will be in the RAW. I think, if I remember correctly, with HTP, LR/ACR are aware of the function being on and increase the exposure(?) to adjust....but not any other curve adjustment as Canon does. Plus, the shadows, potentially, will include more noise.
This is why I turn off all of the "special" functions in the custom features.
-
I should have kept a copy of the original .cr2.
Write to the Library of Congress, you were not the only one duped ;)
-
Also, with the exception of certain raw files (such as Sigma and early .CRW from Canon) when you convert to DNG, everything in the raw file is moved and retained into the DNG even if the DNG Converter doesn't use the metadata.
that is not true if you move back a little... at certain moments Adobe did different decisions about what shall be preserved and what shall not, so if you did not save your original raw files you are stuck w/ what Adobe was thinking is good (big brother knows better) at certain moment (and as they changed their approach that means that their thinking was wrong, hence you lost the data that they think now shall be preserved)... for example for some cameras Adobe DNG converter was discarding part of the data that camera's firmware was writing from masked to light parts of the sensor (just because Adobe's raw conversion had no use for it, but there were valid applications that were using that info).
-
Jeff, thank you so much for helping me work through this. I am going to try and find out about ADL and will post it if I can find anything.
John, do you have any idea where to start looking at how ADL might be decoded or processed by Adobe? My first thought is Adobe Forum, but I have not had much luck there. I suppose I could do an experiment, too. Worst case, there's always Google :P
-
Jeff, thank you so much for helping me work through this. I am going to try and find out about ADL and will post it if I can find anything.
John, do you have any idea where to start looking at how ADL might be decoded or processed by Adobe? My first thought is Adobe Forum, but I have not had much luck there. I suppose I could do an experiment, too. Worst case, there's always Google :P
No, Dave...I'm not sure where I came across the Canon info...Could have been a Canon specific forum, here, DPR, FM forum...??
If I had seen a ADL discussion, I am sure, as a Canon user, I would most likely not focused on it. I would suspect that it would be similar. At worst, you would have a slightly underexposed image in raw. my suggestion, if shooting RAW, would be to turn it off and manage the highlights yourself.
-
do you have any idea where to start looking at how ADL might be decoded or processed by Adobe?
Hi David,
ADL is proprietary to Nikon and if you want your NEF taken with ADL to look like the OOC image when you open it in a Raw converter your only choices are View and Capture NX2. You can get the NEFs to look like that with when opening them with other converters like ACR/LR by putting in a fair amount of work - one good reason to use V/CNX2 is that after all the work in the field to make an image look just right in-camera they save you a fair amount of time by opening them to look just like that. You may want to read through this http://www.luminous-landscape.com/forum/index.php?topic=78788.0 (http://www.luminous-landscape.com/forum/index.php?topic=78788.0) concurrent thread
Cheers,
Jack
-
So I did a little research and found a discussion here (http://photo.net/nikon-camera-forum/00ZTih) that addresses ADL and the Nikon/Adobe differences. But I also decided to do a little experiment with ADL on and off, processed with LR4.4 or ViewNX 2. Bottom line is that there are differences between images converted by the different software. Only the Nikon software is sensitive to ADL set to on or off. In LR4.4, with ADL on or off, the images look the same. So, as Jeff said, LR4.4 doesn't seem to read ADL info. From now on, I will turn ADL off because it interferes a bit with my ETTR thinking and I can better compensate for high contrast scenes by bracketing....as the experts have said here all along! Sometimes, I learn slow :(
For anyone interested in the experimental images, photo filenames starting with _8002393 has ADL ON(normal) and those starting with _8002394 had ADL set to OFF. Below, ViewNX 2 jpeg conversions.
-
Below, the LR4.4 jpeg conversions (no presets)
-
Below, 100% crops of the LR4 conversions, I can't see much of any difference, as expected.
Jeff, I will bet you are rolling your eyes when you read this, but thanks for your patience and for The Digital Negative.
-
Yes, hi David.
You apparently are the lucky owner of two of the world's most advanced cameras, a D800 and a D7100, both with usable Photographic Dynamic Ranges close to 11 stops (http://home.comcast.net/~NikonD70/Charts/PDR.htm#D800), so your need for bracketing is minimized compared to someone who shoots with cameras based on older technology. Add to this the fact that a completely functioning and adapted eye can see only about 12.3 stops of DR (http://www.luminous-landscape.com/forum/index.php?topic=42778.msg633659#msg633659) and one realizes that bracketing with cameras such as yours today is an infrequent exception for special occasions rather than the rule. The same is not true for all current DSLRs, however. Canons, for instance, are typically limited to about 9 stops of PDR.
So then the question for you shifts to a different one: most images are squeezed into 8-bit jpeg files and/or viewed on monitors/prints that can normally render 9 stops or less, with contrast ratios of around 250 or 8 stops being a high norm in practice. So what do we do with the extra couple of stops of DR that recent Nikon DSLRs capture - just let them get lost in the display medium's murky shadows? Of course not. You've got the information, use it at its best. That's where ADL comes into the picture.
From now on, I will turn ADL off because it interferes a bit with my ETTR thinking and I can better compensate for high contrast scenes by bracketing....as the experts have said here all along! Sometimes, I learn slow
If you ETTR in 2013 you probably shoot Raw in full manual mode (http://www.luminous-landscape.com/forum/index.php?topic=78677.msg632709#msg632709), setting shutter speed and f/number according to your artistic costraints of blur, DOF and retained highlights while selecting ISO for the desired balance of SNR, DR and SOOC output brightness. If that's the case ADL does not 'interfere a bit with [your] ETTR thinking'. It doesn't change anything in your Exposure. All it does is intelligently process the 11 stops of your captured PDR to fit into a couple less. It does an excellent job of it. So good that I personally find it quite a time saver in PP to have the NEF look just like that when first opened by the raw converter (CNX2 in my case). Not to mention the much better looking SOOC jpegs.
If you do not shoot in full manual mode, ADL is even smarter than that, but that's the subject for a different post. So don't be afraid of ADL, it's just another advanced tool in your photographer's toolbox. An excellent one at that, so study it and learn how to use it to your advantage, if at all.
Cheers,
Jack
PS Perhaps your other sources shoot with lower PDR cameras? :)
-
If you ETTR in 2013 you probably shoot Raw in full manual mode (http://www.luminous-landscape.com/forum/index.php?topic=78677.msg632709#msg632709), setting shutter speed and f/number according to your artistic costraints of blur, DOF and retained highlights while selecting ISO for the desired balance of SNR, DR and SOOC output brightness. If that's the case ADL does not 'interfere a bit with [your] ETTR thinking'. It doesn't change anything in your Exposure. All it does is intelligently process the 11 stops of your captured PDR to fit into a couple less. It does an excellent job of it. So good that I personally find it quite a time saver in PP to have the NEF look just like that when first opened by the raw converter (CNX2 in my case). Not to mention the much better looking SOOC jpegs.
If you do not shoot in full manual mode, ADL is even smarter than that, but that's the subject for a different post. So don't be afraid of ADL, it's just another advanced tool in your photographer's toolbox. An excellent one at that, so study it and learn how to use it to your advantage, if at all.
Cheers,
Jack
PS Perhaps your other sources shoot with lower PDR cameras? :)
jack, is what you say....ADL not interferring....true if one does not use Nikon raw converters?
I (think) I know what Canon does under the covers with HTP, what does ADL do? How does LR/ACR handle it?
John
-
The only effect for the RAW data in nikon NEF files if you use ADL is perhaps underexposure and this depends on the ADL level (low, high etc). I say perhaps because if you use the low setting it usually do not even change exposure.
This means that if you use ACR/LR, the only thing that you might have achieved by using ADL would be preserved highlights, nothing more. It would be just like using exposure compensation (In manual mode it just change the meter reading, again like using exposure compensation).
Nikon converters apply a propietary algorithm to images with ADL on, I presume it is more complex than just a curve, it looks a more advanced shadow/highlights adjustment.
You can achieve very good results with LR, especially with the new process 2012, playing with the whites/highlights/shadows/blacks sliders.
Going back to the OP, data such as focus point is not used by other converters, and you cannot use the Dust Reference photo either with any other software than Nikon.
-
jack, is what you say....ADL not interferring....true if one does not use Nikon raw converters?
I (think) I know what Canon does under the covers with HTP, what does ADL do? How does LR/ACR handle it?
In full manual metering mode ADL is just an advanced in-camera image processing algorithm applied to the captured Raw data in order to produce the SOOC image/jpeg (Nikon Raw converters use the same algorithm when first opening a NEF captured with ADL on): being smart, it does things like brightness/contrast corrections and tone mapping to obtain what it thinks is a pleasing OOC image given the captured Raw data. In my opinion, for an automatic algorithm, it does an excellent job right out of the post. It does not mess with Exposure - in full manual mode*.
Non Nikon Raw converters start with the exact same Raw data in the NEF, but apply their own camera-specific yet generic recipes which tend to mimic Picture Controls (i.e. Lanscape, Standard, Portrait etc.). But they do not use the information contained in the NEF connected to ADL nor do they attempt to do what it does. Since ADL applies quite a number of adjustments, images taken with ADL on may appear quite different SOOC as opposed to as rendered from a third party Raw converter. I like what it does so I keep it on 90% of the time (I turn it off for instance when I use flash).
Jack
*We need a separate post to deal with ADL used in conjuntion with an automatic metering mode. See below.
-
The only effect for the RAW data in nikon NEF files if you use ADL is perhaps underexposure and this depends on the ADL level (low, high etc). I say perhaps because if you use the low setting it usually do not even change exposure.
Alright so now let's talk about what ADL does in conjunction with an automatic metering mode (NOT in full manual mode, which was discussed in the previous posts). With all this talk about ETTRv2013, wouldn't you like it if your camera Exposed for maximum information captured (i.e. maximized shutter speed, aperture, SNR and DR without clipping) but was also capable of processing that Raw information for a most pleasing brightness/tonality in the final image? You guessed it, that's exactly what auto ADL does when you let it.
When auto ADL is used with Matrix Metering in P,A,S (+ auto ISO) mode for instance, it will evaluate the scene continuously as long as the meter is on before you squeeze the trigger. If it figures that your histogram is going to be empty to the right, it will adjust exposure automatically up to about 1 stop to ETTR. If it figures that your histogram is going to be too full to the right and you might be clipping some desirable highlights, it will adjust exposure automatically up to about 1 stop to shift the histogram to the left in order to let more highlights in. And thanks to current cameras' outstanding PDR it can do it without worries of 'underexposure'. That's how auto ADL determines Exposure to capture the best information (IQ) possible in the Raw data, the NEF.
Next, when you squeeze the trigger, in the blink of an eye it (and V/CNX2) also processes that excellent IQ Raw data into a SOOC image with pleasing brightness and tones as described earlier. If you now open the NEF in a third party converter which does not read ADL information, the rendered image may look off (too bright, too dark, too contrasty or whatever) because they simply apply their generic recipes as is. It doesn't mean that you cannot get an image to look like the one OOC, but you have to work at it. Work that I prefer not to do if I do not have to - because it's already been done for me :)
Jack