Pages: [1]   Go Down

Author Topic: Linear RIMM RGB V4 Profile not linear?  (Read 6248 times)

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Linear RIMM RGB V4 Profile not linear?
« on: November 14, 2013, 11:54:01 am »

For specialized work it is often desirable to work with scene referred data (linear, with no gamma encoding). The ICC web site has a post suggesting how this may be accomplished with Photoshop CS3 and ACR. CS3 is old and ACR now has a new default process (PV2012), but the basics of color management have not changed. However, PV2012 uses adaptive processing and it is difficult to obtain linear data with that version. However, one can use the older process (PV2010) with the suggested workflow. In an attempt to obtain a scene referred image of a Stouffer step wedge, I photographed the wedge with the Nikon D800e and rendered into ProPhotoRGB with linear settings (sliders on the main tab set to zero and the point curve set to linear). I then converted from ProPhotoRGB to LinearRimmRGB ver 4. However, the results are not linear, but have a gamma of 1.18 as shown by analysis with Imatest.



The Imatest log-log plot is linear since such a plot linearizes a power curve, and the shape of the curve can be shown with a linear plot.



To obtain linear data, I created a custom space in Photoshop using ProPhotoRGB primaries but with a gamma of 1.0 and then converted the ProPhotoRGB file to that space, and obtained linear data as shown below. The gamma is 1/1.03.





Bill
Logged

crames

  • Full Member
  • ***
  • Offline Offline
  • Posts: 210
    • http://sites.google.com/site/clifframes/
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #1 on: November 14, 2013, 01:40:48 pm »

Hi Bill,

The linear_RIMM-RGB_v4.icc seems to be linear. If I take an image that has been converted to the linear ProPhoto space, then in PS Assign Profile the linear_RIMM-RGB_v4.icc, I can't see any change, toggling back and forth between the two profiles. I can see a change if I change the gamma a small amount with the Exposure tool.

If I take the image in linear or nonlinear ProPhoto space and Convert to Profile to the linear_RIMM-RGB_v4.icc profile, I see a change in the image when I Convert to Profile using either the Saturation or the Perceptual intent, and further changes when toggling "use black point compensation" on and off. I don't notice a change converting with the Relative Colorimetric intent, nor when toggling black point with that intent.

Since intent and black point compensation might be causing the non-linearities you observe, what are your settings when you convert to linear_RIMM-RGB_v4.icc?

Cliff
Logged
Cliff

Vladimirovich

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1311
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #2 on: November 14, 2013, 05:18:54 pm »

For specialized work it is often desirable to work with scene referred data (linear, with no gamma encoding). The ICC web site has a post suggesting how this may be accomplished with Photoshop CS3 and ACR. CS3 is old and ACR now has a new default process (PV2012), but the basics of color management have not changed. However, PV2012 uses adaptive processing and it is difficult to obtain linear data with that version. However, one can use the older process (PV2010) with the suggested workflow. In an attempt to obtain a scene referred image of a Stouffer step wedge, I photographed the wedge with the Nikon D800e and rendered into ProPhotoRGB with linear settings (sliders on the main tab set to zero and the point curve set to linear). I then converted from ProPhotoRGB to LinearRimmRGB ver 4. However, the results are not linear, but have a gamma of 1.18 as shown by analysis with Imatest.
what about .dcp profile used for conversion... was it w/o nonlinearites introduced by a color transform from scene ref'd data into the working colorspace ?
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #3 on: November 14, 2013, 06:01:12 pm »


Since intent and black point compensation might be causing the non-linearities you observe, what are your settings when you convert to linear_RIMM-RGB_v4.icc?

Cliff

Cliff,

I used relative colorimetric without black point compensation.

Bill

Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #4 on: November 14, 2013, 06:03:37 pm »

what about .dcp profile used for conversion... was it w/o nonlinearites introduced by a color transform from scene ref'd data into the working colorspace ?

Since the ProPhotoRGB file was linear with a gamma of 1.0 prior to the conversion, I do not think the nonlinearity was introduced by the .dcp profile.

Bill
Logged

crames

  • Full Member
  • ***
  • Offline Offline
  • Posts: 210
    • http://sites.google.com/site/clifframes/
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #5 on: November 14, 2013, 07:58:40 pm »

That's strange. Trying to duplicate your workflow by:

1. Take 2 copies of an image in standard ProPhoto RGB,
2. Convert (Relative Intent, no black point) one copy to Linear ProPhoto,
3. Convert the other copy to Linear Rimm.

If I then compare using Apply Image in subtract mode, there is no difference between the two conversions.

Also, converting from Linear ProPhoto to Linear Rimm there is no change, showing no difference in gamma.

I used a synthetic gray scale image containing all RGB values from 0 to 32768, also the Bruce Lindbloom RGB16Million.tif.

I feel certain that the Linear Rimm profile really is linear (with Relative Intent). The non-linearity must be coming from somewhere else. Can you make the NEF and the ProPhoto RGB ACR output available?

Cliff
Logged
Cliff

Vladimirovich

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1311
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #6 on: November 14, 2013, 08:12:20 pm »

Since the ProPhotoRGB file was linear with a gamma of 1.0 prior to the conversion, I do not think the nonlinearity was introduced by the .dcp profile.

Bill

you said "I photographed the wedge with the Nikon D800e and rendered into ProPhotoRGB with linear settings (sliders on the main tab set to zero and the point curve set to linear)." - but did you actually test that data in your file was "linear" ?
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #7 on: November 15, 2013, 05:11:54 am »

you said "I photographed the wedge with the Nikon D800e and rendered into ProPhotoRGB with linear settings (sliders on the main tab set to zero and the point curve set to linear)." - but did you actually test that data in your file was "linear" ?

Hi,

It will usually be affected by veiling glare to some extent. I don't have enough data to say how much was involved in Bill's stepwedge capture. Even with masking of the backlit step-wedge, there will be some spill-over from the lighter regions of the target which will accumulate relatively more in the dark parts, but distance between bright and dark also matters, as does lens design (coating, number of lensgroups, internal light-traps, and blackening of lens edges) and in camera and mirror-box reflections and leaks (I also always block the viewfinder for such tests).

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #8 on: November 15, 2013, 07:31:06 am »

you said "I photographed the wedge with the Nikon D800e and rendered into ProPhotoRGB with linear settings (sliders on the main tab set to zero and the point curve set to linear)." - but did you actually test that data in your file was "linear" ?

Yes, I did, using Rawdigger.

Here is the Rawdigger screen with full stops selected:


And here are the results:


And the graph from Excel:


Regards,

Bill
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #9 on: November 15, 2013, 08:28:17 am »


I feel certain that the Linear Rimm profile really is linear (with Relative Intent). The non-linearity must be coming from somewhere else. Can you make the NEF and the ProPhoto RGB ACR output available?

Cliff

Cliff,

Here is a link to the NEF.

And to the ProPhotoRGB image.

Post edited to show log-log plot of the Green1 channel of the D800e raw file. The upward bend in the shadows is largely veiling flare (I think).


Bill
« Last Edit: November 15, 2013, 09:52:01 am by bjanes »
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #10 on: November 15, 2013, 08:36:32 am »

Hi,

It will usually be affected by veiling glare to some extent. I don't have enough data to say how much was involved in Bill's stepwedge capture. Even with masking of the backlit step-wedge, there will be some spill-over from the lighter regions of the target which will accumulate relatively more in the dark parts, but distance between bright and dark also matters, as does lens design (coating, number of lensgroups, internal light-traps, and blackening of lens edges) and in camera and mirror-box reflections and leaks (I also always block the viewfinder for such tests).

There will always be some veiling glare as Bart noted. In the current case, it was minimized by masking off the surround of the light box and using a relatively simple lens with the latest coatings (the Micro Nikkor 60 mm f/2.8 AFS).

Veiling glare will lift the shadows, which will be best seen in a log-log plot that shows them to advantage.

Here is an image with a lot of veiling glare.



The log-log plot is shown below. The shadows are lifted to give a concave upward curve, which is the opposite of that seen in the LinearRimm image.


Post edited to show log-log plot of the Green1 channel of the D800e raw file. The upward bend in the shadows is largely veiling flare (I think).



Regards,

Bill

« Last Edit: November 15, 2013, 09:53:49 am by bjanes »
Logged

Vladimirovich

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1311
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #11 on: November 15, 2013, 09:58:34 am »

Hi,

It will usually be affected by veiling glare to some extent.

sure, sorry to be PITA - but I was asking about the resulting .TIFF file... output to a colorspace w/ g=1 does not mean that data after color transform (involving Adobe's LUTs and/or embedded tone curve - that depends on profile)/other operations in ACR/LR is linear itself, right ?
« Last Edit: November 15, 2013, 10:01:20 am by Vladimirovich »
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #12 on: November 15, 2013, 10:52:58 am »

sure, sorry to be PITA - but I was asking about the resulting .TIFF file... output to a colorspace w/ g=1 does not mean that data after color transform (involving Adobe's LUTs and/or embedded tone curve - that depends on profile)/other operations in ACR/LR is linear itself, right ?

Yes, the file rendered into ProPhotoRGB with PV2010 and linear settings is linear as expected. Of course, it is encoded with a gamma of 1.8 and is not scene referred.

Imatest LogLog plot:



linear plot using Imatest data:



I hope this finally addresses your concerns.

Regards,

Bill
Logged

crames

  • Full Member
  • ***
  • Offline Offline
  • Posts: 210
    • http://sites.google.com/site/clifframes/
Re: Linear RIMM RGB V4 Profile not linear?
« Reply #13 on: November 16, 2013, 11:11:48 am »

I'm stumped. Your NEF is linear (with small offset/flare) and the ProPhoto RGB file is a clean gamma 1.8.

When I convert your ProPhoto RGB file to Linear ProPhoto and to Linear RIMM, the converted files are essentially identical. I don't see the difference in gamma between the two conversions that you are getting.

Here are the converted files:

Converted to Linear Rimm
Converted to Linear ProPhoto

The only things I can think of is that either, something is happening during profile conversion, or it has to do with Imatest. Are the Imatest Density Response results the same if you look at a single channel, say green, instead of combining the channels?
Logged
Cliff
Pages: [1]   Go Up