Ad
Ad
Ad
Pages: « 1 [2]   Bottom of Page
Print
Author Topic: Can LR really export an original raw file?  (Read 9900 times)
cmi
Sr. Member
****
Offline Offline

Posts: 491


« Reply #20 on: June 29, 2009, 04:39:23 PM »
ReplyReply

Quote from: Schewe
And of course, the question of getting it "right" ain't such an easy task. If you are talking about getting a more accurate rendering of a raw capture, I suspect you would be surprised to find that Adobe Standard with most cameras renders colors more accurately than DP or NX. No, ACR/LR doesn't match the camera jpeg or LCD as easy out of the box. But the camera jpeg or LCD rendering is wrong far more than it's accurate. The camera maker's aren't after "right" they are after a pleasing "look" that makes the image appear better and hides some of their lack of image quality.

If you are thinking somehow that the camera makers know what's "right" I've got some ocean front property I'll sell you in Arizona. Neither Nikon nor Canon have any long term culture or tradition regarding scene rendering. Prior to digital, their primary responsibility was to maintain a light tight environment and to form a sharp image the film. When they had to come up with image processing to take raw captures into jpegs they adopted existing technologies and used their own image expectations regarding what the image should look like and "right" wasn't really high on their priority list ya know.

If by "right" you mean the first two dimensional representation (you camera's LCD) you really, seriously need to quit chimping so much and not fall in love with scifi (which is what the back of the camera is showing you).

I see you have no inhibitions to object to very direct language!   You seem to be used defending yourself frequently from harsh critique!

I read my text again and see how you could come to the impression Im after the Cameras default rendering. Of course I tweak individually and am not after the cameras default. Was so self-evident for me that I didnt mentioned it. So, if I tweak a series of images both in LR and DPP, and I do this frequently, generally I like the tonal result of DPP more often then the look from LR/ACR (with camera profiles.) Period. My experience, Im just mentioning it.

And thanks for your explanation why it is so! So ACR/LR deliberately does not try to sweeten and to fix. Was not thinking about this before, but makes perfect sense. I would not see fixing and sweetening as bad per se, when I judge from my results, but I can see that it would be extremely hard if not impossible to implement (for each camera). So the only chance IS to make it as accurate as possible, with the consequence that individual rendering MUST fall a bit short, wich you adressed partly with custom curves. At least thats how I understand it now. Again, thank you very much!

Christian


//Edit: Do you see it as bad that the other vendors implement sweeteners/fixes? You sound to me a bit like that. (Ok they should allow to turn it on/off.)
« Last Edit: June 29, 2009, 05:02:46 PM by Christian Miersch » Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5542


WWW
« Reply #21 on: June 30, 2009, 02:10:47 AM »
ReplyReply

Quote from: Christian Miersch
//Edit: Do you see it as bad that the other vendors implement sweeteners/fixes? You sound to me a bit like that. (Ok they should allow to turn it on/off.)


Yes, I think it sucks because it leads people to expect or want essentially inaccurate results and then somehow get it in their head that the sweetened rendering is the correct one.

I go back to film days and bet that chrome shooters tend to like the camera JPEG rendering because it's got a steepened tone curve and punched up colors like Kodachrome. People who used to shoot B&W or color neg prefer to see an accurate rendering of what's in the file and don't want artificial tone curves or saturation added.

People who prefer to season to taste rather than have the chef dictate the seasoning don't have a problem tuning an image to render they way they want...
Logged
cmi
Sr. Member
****
Offline Offline

Posts: 491


« Reply #22 on: June 30, 2009, 04:37:37 AM »
ReplyReply

Thanks Jeff,

I thought so. My opinion differs in some aspects. But - getting to know the LR/ACR philosophy from you was interesting, was very insightful for me.

best regards

Christian


Quote
Yes, I think it sucks because it leads people to expect or want essentially inaccurate results and then somehow get it in their head that the sweetened rendering is the correct one.
Logged
cmi
Sr. Member
****
Offline Offline

Posts: 491


« Reply #23 on: June 30, 2009, 06:08:28 AM »
ReplyReply

I know its OT, but the issue kept my busy thinking. So here is how I see it now:

If a fix adressed a technical weakness and could be turned on/off, then it would be ok. Same for sweetening. But since it cannot be controlled in the vendors softwares, people dont recognize these fixes, and that leads to wrong expectations. Lightroom on the other hand just cant emulate everything 100%. Likely because these fixes are likely not documented, and it is also the question if they are always good.

So from a user perspective this creates the irritating result that quite often DPP looks better - but, sometimes, I instead prefer the Lightroom/ACR version. Either this are the cases where these fixes fail, or where I happened to just tweak one version better. Really hard to know, since the fixes cant be separated. At least I dont know a way.

Correct?
« Last Edit: June 30, 2009, 06:11:08 AM by Christian Miersch » Logged
pegelli
Sr. Member
****
Offline Offline

Posts: 609



WWW
« Reply #24 on: July 01, 2009, 12:31:50 AM »
ReplyReply

Quote from: Christian Miersch
I know its OT, but the issue kept my busy thinking. So here is how I see it now:

If a fix adressed a technical weakness and could be turned on/off, then it would be ok. Same for sweetening. But since it cannot be controlled in the vendors softwares, people dont recognize these fixes, and that leads to wrong expectations. Lightroom on the other hand just cant emulate everything 100%. Likely because these fixes are likely not documented, and it is also the question if they are always good.

So from a user perspective this creates the irritating result that quite often DPP looks better - but, sometimes, I instead prefer the Lightroom/ACR version. Either this are the cases where these fixes fail, or where I happened to just tweak one version better. Really hard to know, since the fixes cant be separated. At least I dont know a way.

Correct?

Interesting perspective, don't know if this is why I have taken the path I have which is:

1: Used a color checker to determine calibration slider settings in lightroom such that my two DSLR bodies give very close standard conversion results of the same scene (even though their technology vintage is about 3 years apart)
2: Only use lightroom for conversion and really try to learn that well

I think in the end that will lead to more consistent results between my pictures which in my mind is more valuable than switching software/converters and spend endless times trying to create similar looks between the different workflows.
Yes, once in a while after reading online forums I have a nagging feeling that maybe noise or color could be improved for some pictures using a new approach, but in the end I rather spend that time taking pictures or becoming better in my standard workflow as there the return on time investment is greater. This perspective is heavily influenced by being a pure amateur photographer with limited time to spend on a hobby. This approach will obviously not work for everybody, but for me it has given me great pleasure and a continuous move up the learning curve with still many miles to go.
Logged

pieter, aka pegelli
simdoc1
Newbie
*
Offline Offline

Posts: 34


« Reply #25 on: July 04, 2009, 07:04:54 AM »
ReplyReply

Quote from: Schewe
If you are thinking somehow that the camera makers know what's "right" I've got some ocean front property I'll sell you in Arizona. Neither Nikon nor Canon have any long term culture or tradition regarding scene rendering. Prior to digital, their primary responsibility was to maintain a light tight environment and to form a sharp image the film. When they had to come up with image processing to take raw captures into jpegs they adopted existing technologies and used their own image expectations regarding what the image should look like and "right" wasn't really high on their priority list ya know.

I am late to this party, but I am quite struck by this concept of "right" - especially given how often it is the topic of adversarial proceedings on this and other forums. I had never thought of how it was that Nikon, Canon, etc. became the custodians of "right"; however, of course, this prerogative used to belong to Kodak, Fuji, Agfa and others. The best example of this are the film emulations which are part of many software packages - each one aiming to produce a version of "right" based on a different standard. In that context, a RAW conversion that is "accurate" and technically sound is an incredible value because it gives you a reliable starting point for moving towards your idea of "right" - whatever that may be.
Logged
milt
Jr. Member
**
Offline Offline

Posts: 70



WWW
« Reply #26 on: July 07, 2009, 07:57:01 PM »
ReplyReply

Quote from: Schewe
No...it wrote into the .xmp file (or the xmp bock in the case of DNG) only. If you toss the .xmp file the raw file is still absolutely untouched...what the bug was causing was problems in the database record of the file...not the original file at all.

You may have been confused abut an earlier version of Capture NX actually writing inside the actual raw file and that bug resulted in the raw NEF being corrupted...

Jeff,

I'm sorry but, as the result of some research, I must respectfully disagree.

First, for LR version 1, here is a direct quote from Victoria Brampton's "Missing FAQ" book, page 164:

"In version 1, for proprietary raw files, most metadata is written to an XMP sidecar file, however, the updated capture time is written back to the raw file itself, changing the modified date of the file."

"That shouldn't cause any problems as it's only written to the metadata file header, and doesn't affect the raw image data itself, however it is a point of concern for some people who they[sic] feel that the raw files should never be touched in any way."

As to the bug in version 1, here is forum thread discussing it:

http://forums.adobe.com/message/1454773#1454773

Now, for LR version 2, it looks like there are two modes.  The next paragraph in Victoria's book is:

"For that reason, in version 2, there is a setting in Catalog Settings > Metadata panel to allow you to choose whether the updated time is stored only in the catalog, exported files, and XMP sidecar files, or whether it can be updated in your original raw files."

So, considering Michael's original statement of:

"Lightroom never touches the original raw file. All edits are virtual."

I think a fairer statement would have been:

"For LR V2, the original raw file is never touched, with the exception of the case of changing the metadata capture date of the photo when you have the appropriate mode set in the Metadata panel.  (For LR V1, changing the metadata capture date of the photo always changes the original raw file.)"

--Milt--
Logged

Los Gatos, California | http://miltonbarber.com
Schewe
Sr. Member
****
Offline Offline

Posts: 5542


WWW
« Reply #27 on: July 08, 2009, 01:03:58 AM »
ReplyReply

Quote from: milt
"For LR V2, the original raw file is never touched, with the exception of the case of changing the metadata capture date of the photo when you have the appropriate mode set in the Metadata panel.  (For LR V1, changing the metadata capture date of the photo always changes the original raw file.)"


Yeah, well I actually set my cameras to the correct date and time when I start shooting an important project–it's one of my check list items...and as you note there is an option to alter or not alter the time stamp in 2.x. Me, personally, I really wouldn't want to touch original the raw file (proprietary raw)  and only edit DNGs that are designed for it.
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2882



« Reply #28 on: July 08, 2009, 02:01:40 PM »
ReplyReply

Quote from: Schewe
Yes, I think it sucks because it leads people to expect or want essentially inaccurate results and then somehow get it in their head that the sweetened rendering is the correct one.

I go back to film days and bet that chrome shooters tend to like the camera JPEG rendering because it's got a steepened tone curve and punched up colors like Kodachrome. People who used to shoot B&W or color neg prefer to see an accurate rendering of what's in the file and don't want artificial tone curves or saturation added.

People who prefer to season to taste rather than have the chef dictate the seasoning don't have a problem tuning an image to render they way they want...

If Mr. Schewe is interested in an "accurate" representation of the scene, then he should consider using scene referred data, where the recorded pixel value is proportional to the scene luminance. A raw file is scene referred within the dynamic range response of the sensor. One can easily obtain scene referred data with ACR as described in an ICC White Paper. The first step is to set the sliders on the main tab of ACR to zero (contrast, brightness, black point, linear tone curve, etc) and convert the raw file. This will essentially give scene referred data with a gamma encoding; the gamma encoding can be removed by converting to a linear profile if desired. The problem with scene referred data is that the dynamic range usually exceeds that of the display device. To obtain a useful display or print, the data must be compressed so that they fit the dynamic range of the output device. This is done by rendering, as discussed in an Adobe white paper by Karl Lang. This process usually involves rolling off the highlights and shadows and increasing contrast in the midtones, essentially applying an "S curve" to the data.

Both ACR and the JPEG engine of the camera render the images as described above, but the default curves vary somewhat. For example, here are characteristic curves obtained with ACR and the Nikon D3:

[attachment=15222:TRCsBl.gif]

The ACR defaults for the D3 have a black setting of 5, which rolls off the shadows, but the standard camera curve does not. If one sets the black to zero with the Adobe standard curve, the results are similar to the Camera Standard curve, but the camera standard curve has a steeper slope for the midtones. The Camera Standard curve is very similar to what would be obtained with the camera JPEG engine set to the Standard Picture Control. Both the Adobe and Camera TRCs boost midtone contrast and neither are accurate. If one wants "accurate" data, the ACR linear tone curve could be used, but the results are not pleasing. Some photographers (such as Mark Segal, an expert who contributes articles to this site) like to start with a linear curve so as to see what is in the image and then apply the necessary edits. Others prefer to start out with an image that more closely resembles the final image. As long as you know what you are actually doing, these approaches are matters of personal preference.

Anyone who has looked at the H & D curve of a black and white negative film knows that it is not linear. It has a knee and shoulder just like the Adobe Standard curve or the Nikon curve. In the first case, the curve is applied automatically by the process. Also, the slope of the curve is less than unity and is affected by development. Mr. Schewe is not looking at "accurate data" in his black and white negative. Scene referred data may be extracted from film, but the process is not simple (see the paper by Debevec).
Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5542


WWW
« Reply #29 on: July 08, 2009, 04:37:38 PM »
ReplyReply

Quote from: bjanes
Anyone who has looked at the H & D curve of a black and white negative film knows that it is not linear. It has a knee and shoulder...


Huh...and here all these years I thought it was called a TOE and shoulder. Leastways that's what Minor White called in back in school at RIT.
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2882



« Reply #30 on: July 08, 2009, 05:26:47 PM »
ReplyReply

Quote from: Schewe
Huh...and here all these years I thought it was called a TOE and shoulder. Leastways that's what Minor White called in back in school at RIT.

I stand corrected, it is TOE not knee. I misspoke extemporaneously, but the substance of my argument is correct (unlike yours). Do you care to comment on "accuracy"?
Logged
ErikKaffehr
Sr. Member
****
Online Online

Posts: 8023


WWW
« Reply #31 on: July 09, 2009, 12:27:49 AM »
ReplyReply

Hi,

To my best understanding the linear DNG:s from DxO are demosaiced. In my view this is a major issue because DxO's demosaic algorithm may not be the best one. Also, keeping original, non demosaiced, non interpolated RAW makes it possible to use different algorithms in the future. So going to "linear" DNG looses some information.

When doing demosaicing the file size increases by a factor of three, as two new pixels are interpolated for each pixel. So big file means that some spacial information is lost.

I'm not very happy about the state of "RAW". We need something like DNG, IMHO. Unfortunately some RAW-converters, namely DxO and Bibble, don't support DNG, at least as a true RAW format.

Lightroom lacks some essential features IMHO.

- It would be nice if we had a "heavy weight" noise reduction like NoiseNinja integated in Lightroom as a parametric edit.
- Lens corrections would be needed
- Perspective correction would be needed

The reason is really that the whole idea with Lightroom is parametric edits. Once you leave Lightroom trough an external editor you are actually working on a rendered file.

The other issue is that especially noise reduction should be done before sharpening. Lightroom has AFAIK pretty good capture sharpening, but you should probably not use with Noise Ninja, Dfine or the other tools.


Best regards
Erik

Quote from: Schewe
When Capture One 4.8.x outputs a DNG, it does the actual demosiacing....so any lens corrections (and AtoD corrections) are performed on the linear DNG output. In that regard it's much like DxO's linear DNG lens corrections...

You do need the demosiacing to take advantage of the lens corrections but saving out the linear DNG means you aren't locked into Capture One's color/tone output to a tiff file. The DNG's ain't tiny mind you...I think they're about 117MB and it does take a degree of zen like patients when working on the files in ACR/LR...but I much prefer ACR/LR tone and color controls compared to Capture One (not to mention the image sharpening).

:~)
Logged

Schewe
Sr. Member
****
Offline Offline

Posts: 5542


WWW
« Reply #32 on: July 09, 2009, 01:04:08 AM »
ReplyReply

Quote from: ErikKaffehr
Lightroom lacks some essential features IMHO.

- It would be nice if we had a "heavy weight" noise reduction like NoiseNinja integated in Lightroom as a parametric edit.
- Lens corrections would be needed
- Perspective correction would be needed

Don't be at all surprised if the Camera Raw/Lightroom elves have a little something up their sleeves with regards to your wishes...

:~)

And I really only suggested Capture One linear DNG as a workaround to the lack of lens corrections in Camera Raw/Lightroom. It gives me the lens corrections of Capture One but allows me to keep using the color and tone correction that I prefer in ACR/LR. But I don't see that as a long term solution...linear DNGs from a P 65+ back are pretty figgin' big, ya know?
Logged
cmi
Sr. Member
****
Offline Offline

Posts: 491


« Reply #33 on: July 09, 2009, 08:20:55 AM »
ReplyReply

Quote from: bjanes
... "accurate" representation of the scene...

Thanks for your post and the mention of the papers. I was wondering about that "accurate" as well.
Logged
Pages: « 1 [2]   Top of Page
Print
Jump to:  

Ad
Ad
Ad