Pages: 1 2 3 [4] 5   Go Down

Author Topic: More Color Checker questions WRT accurate color  (Read 42367 times)

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: More Color Checker questions WRT accurate color
« Reply #60 on: November 30, 2011, 10:30:28 am »

Actually, not. My understanding is that while DNG converter is smart enough to understand all the raw formats that ACR/LR understands, and all of the color rendering and lens correction that can be encoded into a DNG camera profile, the demosaicing, noise reduction, etc algorithms are not as advanced as ACR/LR - they're just good enough to be able to create thumbnails and previews. So you can't in effect use the DNG converter as a no-cost substitute for LR/ACR

Sandy

certainly it creates just highly compressed JPGs and not 16bit TIFFs (= "So you can't in effect use the DNG converter as a no-cost substitute for LR/ACR"), but think again - why 'd Adobe developers maintain a separate code for that part of DNG converter functionality vs ACR/LR  ;D ? the code in Adobe DNG converter even supports all the local edits w/ brush that you can do in ACR/LR...

Logged

sandymc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 350
Re: More Color Checker questions WRT accurate color
« Reply #61 on: November 30, 2011, 10:40:36 am »

certainly it creates just highly compressed JPGs and not 16bit TIFFs (= "So you can't in effect use the DNG converter as a no-cost substitute for LR/ACR"), but think again - why 'd Adobe developers maintain a separate code for that part of DNG converter functionality vs ACR/LR  ;D ? the code in Adobe DNG converter even supports all the local edits w/ brush that you can do in ACR/LR...

They don't need to - there's a low performance (very very low  ;D) demosaicing module in the DNG SDK. Which sure isn't what's in ACR/LR.

Sandy
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: More Color Checker questions WRT accurate color
« Reply #62 on: November 30, 2011, 10:55:33 am »

They don't need to - there's a low performance (very very low  ;D) demosaicing module in the DNG SDK. Which sure isn't what's in ACR/LR.

Sandy

well, we can ask Eric Chan directly or we can do a test... let us see... new sharpening technology in recent versions of ACR uses some form of deconvolution, so we can push the sliders to make it prominent, run DNG converter and see if the JPG thumbnail will display the same effect as seen in ACR... OK ? and we can also test NR the same way - push sliders so that effect in ACR will be very-very noticeable to everybody's eye and then check if the same is seen in JPG thumbnail... that low performance demosaicing module - has it any NR there at all  ;) ?
« Last Edit: November 30, 2011, 10:59:59 am by deejjjaaaa »
Logged

RFPhotography

  • Guest
Re: More Color Checker questions WRT accurate color
« Reply #63 on: November 30, 2011, 12:01:42 pm »

Which, in the end, is all essentially moot anyway.  The DNG PE backs out edits made in LR/ACR except for WB.  The only thing I wasn't sure of was whether the same default tone curve was applied via the DNG Converter vs via an export out of LR/ACR.
Logged

sandymc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 350
Re: More Color Checker questions WRT accurate color
« Reply #64 on: November 30, 2011, 12:31:16 pm »

well, we can ask Eric Chan directly or we can do a test... let us see... new sharpening technology in recent versions of ACR uses some form of deconvolution, so we can push the sliders to make it prominent, run DNG converter and see if the JPG thumbnail will display the same effect as seen in ACR... OK ? and we can also test NR the same way - push sliders so that effect in ACR will be very-very noticeable to everybody's eye and then check if the same is seen in JPG thumbnail... that low performance demosaicing module - has it any NR there at all  ;) ?

Well, I did ask Eric a while ago, and he said that's what's in the DNG converter isn't what's in ACR/LR. But's that quite a long time ago, so things may have changed.

Sandy
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: More Color Checker questions WRT accurate color
« Reply #65 on: November 30, 2011, 02:34:30 pm »

Well, I did ask Eric a while ago, and he said that's what's in the DNG converter isn't what's in ACR/LR. But's that quite a long time ago, so things may have changed.

Sandy

I think I remember that topic, but what is the exact quote of your question/his answer... let me try to find it...

here it is :

"...Hi Sandy, yes, you are right: the DNG SDK only has a simple bilinear interpolation demosaic routine implemented. (We have been meaning to update it with something more serviceable, but have not worked that into the already rather-packed dev schedule ..."


Eric is talking about available in source code form SDK, not closed source Adobe DNG Converter... and that is the point - Adobe DNG converter has the full blown ACR/LR raw conversion/adjustment interpretation code inside, DNG SDK being available for a general public as a source code does not have that functionality.
« Last Edit: November 30, 2011, 02:43:50 pm by deejjjaaaa »
Logged

RFPhotography

  • Guest
Re: More Color Checker questions WRT accurate color
« Reply #66 on: November 30, 2011, 04:29:37 pm »

Does this exchange between Eric and Nat Coalson change anyone's opinion on what's in a DNG and what isn't?

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

Schewe

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6229
    • http:www.schewephoto.com
Re: More Color Checker questions WRT accurate color
« Reply #67 on: November 30, 2011, 06:03:54 pm »

Does this exchange between Eric and Nat Coalson change anyone's opinion on what's in a DNG and what isn't?

Nope...it's a standardized file format wrapper, just like Eric said. There are technical issues and political issues, Eric addressed the technical issues. The politics of DNG are considerably murkier than the technical issues because so many people have incorrect assumptions based on faulty technical understanding.
Logged

RFPhotography

  • Guest
Re: More Color Checker questions WRT accurate color
« Reply #68 on: November 30, 2011, 06:35:09 pm »

Nope...it's a standardized file format wrapper, just like Eric said. There are technical issues and political issues, Eric addressed the technical issues. The politics of DNG are considerably murkier than the technical issues because so many people have incorrect assumptions based on faulty technical understanding.

LOL.  There's a shock, Jeff.  I was really posing that question to deejjjaaaa because Eric's explanation seems to conflict with deejjjaaaa's understanding. 

Intuitively, to me, it seems odd that conversion to DNG would apply a default tone curve based on an Adobe profile (e.g., Adobe Standard).  It would seem that would cause a difference in the appearance of two images, one original Raw and one DNG, in any other program that was able to read DNG files, like Photomatix wouldn't it?  I just tried it.  Converted 5 Raw files to DNG, loaded the DNG files into PM, tonemapped and saved then loaded the same 5 Raw files into PM, tonemapped and saved and the two tonemapped files are identical.  I loaded the files into PM directly, not via the LR plugin, so PM is doing the conversion.  Doesn't that indicate that the DNG doesn't have a default Adobe tone curve applied to it?  Exporting those same 5 files out of LR to PM produces a different result.  I've got an import preset that zeroes out everything (incl Blacks, Brightness, Contrast) and sets the curve to linear so the only thing that should be different on export is the application of the Adobe Standard tone curve.  That tonemapped image is different. 

Even within an Adobe program, unless LR or ACR handled DNG files differently from how it handles other files, there'd be a doubling of the tone curve, once for what was embedded in the DNG file and again when the DNG file was opened in LR/ACR based on whatever profile was chosen in the Camera Calibration function.  Wouldn't there?
Logged

Schewe

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6229
    • http:www.schewephoto.com
Re: More Color Checker questions WRT accurate color
« Reply #69 on: November 30, 2011, 07:24:55 pm »

Intuitively, to me, it seems odd that conversion to DNG would apply a default tone curve based on an Adobe profile (e.g., Adobe Standard).

It does...based on the "Default" ACR/LR settings YOU have set as default. If you haven't changed your ACR/LR defaults then they will be the same "Defaults" Thomas and Eric decided should be default and pretty much all other applications that support DNGs would see the raw DNG the same way. And it will currently be with the "medium" tone curve applied–unless you've changed the Defaults...

It's important to note that Camera Raw, DNG Converter and Camera Raw all share a per machine Default settings...if you change your Defaults in ACR or LR then DNG Converter's Defaults will also change. There is no UI to alter the DNG Converter's Defaults though.
Logged

RDoc

  • Newbie
  • *
  • Offline Offline
  • Posts: 18
Re: More Color Checker questions WRT accurate color
« Reply #70 on: November 30, 2011, 07:30:53 pm »

Yes, that's right.  If you build a single-illuminant profile in DNG PE, the target values used by DNG PE are from a set of averaged ColorChecker charts with adopted illuminant of D50.  If you build a dual-illuminant profile in DNG PE, the illuminants used by DNG PE are A and D65.

They should be visually close, yes, at default ACR settings. (But not numerically close. The readouts in ACR are different from the readouts in DNG PE.)

They'll be different numerically & visually from the ideal values because of the default tone curve rendering and black subtraction that ACR applies, which adds contrast and modifies saturation (as Sandy noted, above).  The published ideal values assume no additional rendering is done.  If you want to get a closer match, you would need to turn off the default ACR rendering.  That is done by setting Brightness, Contrast, and Blacks to 0 (instead of 50, 25, and 5) and settings the point curve to Linear instead of Medium Contrast.

Which is basically what I had to do in the posted samples above to get exact numerical Lab values to the published numbers Bjanes posted and finally measured from shown in his Imatest posting.

Except instead of zeroing out all settings, I had to modify a "linearish" tone curve and set Brightness to +60 in ACR. I built the DNG Profile used in that image sample using ACR defaults embedded in the DNG input image.

I've been trying that all along with OK but far from perfect results. If I shut off all the ACR edits the results are very far off from the published values. If I adjust the tone curve to match the gray patch published values, with or without changing the exposure or brightness first to get it close the values are OK but not perfect.

My original question was what was happening that made the profile transform not produce the published values. The assumption being that the DNG created profile is a transform designed to map from an image to the color checker. If that were true then applying that transform to the original image should produce something very close to the values programmed into the DNG editor.

Is the major issue here that because the profile doesn't produce a full tone correction, the transform isn't complete? That is, the DNG PE profile doesn't attempt to really reproduce the color checker in the Prophoto color space because it's not fully dealing with the tone curve? Since attempting to later reproduce the tone curve in ACR is both inexact and modifies all 3 Lab components of all the patches it's pretty much impossible to get back to the published values. Every tweak to the tone of the gray patches is also moving all 3 of the the color patch values around in a non-linear manner.

I did try playing with the tone curve in the DNG Editor to see if it was possible to get the profile to reproduce the tone curve. I started both from the base curve and from linear but both produced quite unexpected results when applied to the image in ACR. It's not clear to me what this is doing or what the input/output numbers for the control points mean, but that's really a separate topic.



Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: More Color Checker questions WRT accurate color
« Reply #71 on: December 01, 2011, 02:24:55 am »

Quote
My original question was what was happening that made the profile transform not produce the published values.

You're expecting too much out of an image editing program designed for photographers who rely on visual judgement in determining what looks good to them. Exact numerical color matching is almost impossible to pull off by most photographers due mostly to the fact they have to rely on the limitations of their memory of the actual scene. Let me reemphasize the Eric Chan quote you posted...

Quote
They should be visually close, yes, at default ACR settings. (But not numerically close. The readouts in ACR are different from the readouts in DNG PE.)

If you want an exact visual match of your painting viewed on your display in ACR/LR to match the actual painting you see with your eyes, you're better off doing this using your eyes to guide your edits along with a custom camera profile. The main purpose of a camera profile using the DNG PE Wizard is to get the hues='a/b' Lab channel relationships to remain the same when applying contrast, brightness which in turn affect saturation levels. Human vision sees hue errors more so over any other of the HSL color channels which is why ACR/LR includes an HSL panel.

Believe me, it's easier than you think to get an accurate reproduction using ACR/LR. I've done it myself countless times photographing objects having a wide range of colors lit by various types of continuous lights, placed next to my display and editing to get an exact match. It's uncanny how close I get, but I have to use not just the profile but the HSL, WB sliders, SplitTone panel to get a match.

Sometimes I can just use Adobe Standard or ACR 4.4 default especially if I use neutral-ish fluorescent lights which I have several different brands. And due to these light's spiky spectra HSL panel and WB sliders really fix a lot of the hue errors.

This is why you need to shoot your paintings under full spectrum light source like Solux 4700K bulbs, strobes or flash. It will make editing a whole lot easier.
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: More Color Checker questions WRT accurate color
« Reply #72 on: December 01, 2011, 02:43:18 am »

Oh, and I found an added bonus converting my Raw PEF's to DNG by saving out of ACR in that format. I went into ACR's (4.6) Preferences and checked both boxes in the DNG file handling section to "ignore xmp files" and "Update Jpeg Previews" set to "Full Size".

Now when I save out of ACR to DNG format I select Jpeg Preview=Full Size and now sharpening shows up in Bridge CS3's Preview Pane of ONLY DNG previews of my Raw PEF's. I could never get my PEFs to show applied sharpening in the image preview even when I have selected "Use High Quality Previews" in Bridge Preferences. I even purged the cache and it doesn't work.

I don't know if choosing Full Size JPEG previews or Update JPEG previews caused this to happen.
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: More Color Checker questions WRT accurate color
« Reply #73 on: December 01, 2011, 03:17:12 am »

based on faulty technical understanding.
based on non disclosure by Adobe
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: More Color Checker questions WRT accurate color
« Reply #74 on: December 01, 2011, 03:26:24 am »

I was really posing that question to deejjjaaaa because Eric's explanation seems to conflict with deejjjaaaa's understanding.
I was only stating that Adobe DNG converter has the same __code__ as ACR/LR to deal w/ raw conversion/sharpening/NR and various adjustment interpretation (like local edits = adjustment brush, etc) and that Adobe DNG converter is perfectly capable to generate JPG output (degraded by compression of course) natuarally matching what ACR/LR will produce (again taking into consideration that it will be a highly compressed JPG intended for preview purposes and stored inside DNG file - but you can extract it from there w/ various tools)... and I did not see anything posted by Eric to contradict that (Adobe DNG converter != DNG SDK).
Logged

sandymc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 350
Re: More Color Checker questions WRT accurate color
« Reply #75 on: December 01, 2011, 05:31:56 am »

Eric is talking about available in source code form SDK, not closed source Adobe DNG Converter... and that is the point - Adobe DNG converter has the full blown ACR/LR raw conversion/adjustment interpretation code inside, DNG SDK being available for a general public as a source code does not have that functionality.

Well, sort of. This is actually the relevant part of my post that Eric was responding to:

Quote
Umm, DNG converter has the full ACR demosaicing algorithm embedded in it, rather than a simplified, lower performance version? That would surprise me. But then, I get surprised on a regular basis.....

But subsequently Eric and I have also had several "off-forum" conversations - while those conversation were not about the DNG converter, I can only repeat that my understanding is that the DNG converter doesn't have the full ACR/LR conversion engine in it. But as I said in the quote above, in the field of color management, I get surprised regularly :o

Sandy

Sandy
Logged

RFPhotography

  • Guest
Re: More Color Checker questions WRT accurate color
« Reply #76 on: December 01, 2011, 07:58:12 am »

It does...based on the "Default" ACR/LR settings YOU have set as default. If you haven't changed your ACR/LR defaults then they will be the same "Defaults" Thomas and Eric decided should be default and pretty much all other applications that support DNGs would see the raw DNG the same way. And it will currently be with the "medium" tone curve applied–unless you've changed the Defaults...

It's important to note that Camera Raw, DNG Converter and Camera Raw all share a per machine Default settings...if you change your Defaults in ACR or LR then DNG Converter's Defaults will also change. There is no UI to alter the DNG Converter's Defaults though.

Yep, I get that, Jeff.  But the one thing I can't change in ACR/LR is the application of a profile via the Camera Calibration pane.  That's the only aspect I'm uncertain about.  Whether it's Adobe Standard, Camera Faithful or whatever one of those Adobe profiles is used, there's a tone curve baked in.  Does that profile get taken into account when converting to a DNG?  If it does then I don't believe that's the best starting point for trying to create a custom DNG camera profile (maybe my belief is wrong).  My thinking is that the best starting point is the linear raw data captured by the camera (or as close to it as possible).  If there's a tone curve in the DNG as a result of the profile used in the Calibration pane (or if there's one baked into the DNG Converter) then it's not the linear camera data anymore.  Now, DNG PE does back out any user edits (except WB) when the DNG is opened in that application.  If PE also backs out the tone curve of the calibration profile, no problem.  But I don't know if it does.  Does it?
« Last Edit: December 01, 2011, 08:04:23 am by BobFisher »
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: More Color Checker questions WRT accurate color
« Reply #77 on: December 01, 2011, 09:53:59 am »

Well, sort of. This is actually the relevant part of my post that Eric was responding to:

1) you asked about Adobe DNG converter (closed source) and Eric answered you about Adobe DNG SDK (open source)...

why ? may be because :

2) the issue discussed there was "original raw file" -> "linear DNG file" conversion, which is different from what we (you and me - not the original theme of the topic) are discussing now (generation of embedded JPG previews).
« Last Edit: December 01, 2011, 09:57:43 am by deejjjaaaa »
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: More Color Checker questions WRT accurate color
« Reply #78 on: December 01, 2011, 10:04:38 am »

then it's not the linear camera data anymore.

some cameras already make the data "non linear" before writing their raw files and then raw converters make the reverse application of a "curve" to the data when reading it.

for example : http://translate.google.com/translate?sl=ru&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&layout=2&eotf=1&u=http%3A%2F%2Fblog.lexa.ru%2F2011%2F11%2F10%2Fo_lineinosti_raw_nikon_d5x00.html
Logged

RFPhotography

  • Guest
Re: More Color Checker questions WRT accurate color
« Reply #79 on: December 01, 2011, 10:31:07 am »

The translation on that link is useless.  But that still doesn't answer the question. 
Logged
Pages: 1 2 3 [4] 5   Go Up