Worth the trouble? Different camera profiles for different ISO's?

[l_d_allan]

My impression is that images of ColorChecker, SpyderCheckr, and/or QpCard targets might vary with major differences in ISO. It seems like the RGB/L*a*b values of the target patches could vary at ISO 100 vs ISO 400 vs ISO 1600 vs ISO 6400, etc. The noise and DR might also vary.

Or not? Would this tend to be too subtle to make a difference? I haven't really checked, and I'm unclear how it could be measured.

If there are differences beyond some threshold, would it be worth the trouble to make different camera profiles for different ISO's?

Note: I'm aware that a persuasive case can be made that camera profiles tend to not be particularly worthwhile, even for reprographics where accuracy counts. "Do it by eyeball so it looks good."

For this question, how about accepting the premise that camera profiles are worthwhile, even for non-reprographics, to get a reasonable baseline starting point for subsequent p.p.?

[D Fosse]

I've never observed any color shifts. Noise and DR obviously, but not color (but then I rarely go above 1600).

I've made a lot of color checker profiles for various Nikons, and in every case decided that the Adobe Standard profile yields more consistently accurate results. No contest, really. With the CC profiles there is always at least one "runaway" color range that oversaturates and/or shifts wildly and uncontrollably. My theory is that there are simply not enough patches (I'm using the passport).

I do have a few profiles for odd (non-spectrally uniform) lighting, fluorescent, LED and so on, but I use them with the provision that they may not be entirely "accurate", just ballpark.

So especially for "reprographics where accuracy counts", I distrust CC profiles. Accurate color is an illusion in any case. I always aim for "equivalent" rather than "accurate". That tends to be accepted as accurate.

[digitaldog]

I can only speak to the process of DNG profiles built with X-Rite's and Adobe's software, there's no reason to worry about ISO (or for that matter, building lots of DNG profiles for each scene). For odd illuminants (Fluorescent, LED) yes, you want to build a custom profile and be done.

This new 30 minute video on DNG camera profiles discusses this and shows the effects of ISO 100 vs. 800 upon the target (spoiler alert, there's none):

DNG camera profile video tutorial
January 2014
In this 30 minute video, we’ll look into the creation and use of DNG camera profiles in three raw converters. The video covers:

What are DNG camera profiles, how do they differ from ICC camera profiles.
Misconceptions about DNG camera profiles.
Just when, and why do you need to build custom DNG camera profiles?
How to build custom DNG camera profiles using the X-rite Passport software.
The role of various illuminants on camera sensors and DNG camera profiles.
Dual Illuminant DNG camera profiles.
Examples of usage of DNG camera profiles in Lightroom, ACR, and Iridient Developer.

Low Rez (YouTube):
http://youtu.be/_fikTm8XIt4

High Rez (download):
http://www.digitaldog.net/files/DNG%20Camera%20profile%20video.mov

[Tim Lookingbill]

Thanks for producing and giving a heads up on your video, Andrew.

It confirms that illuminant specific based profiles don't affect colors the same way across all camera brands/models when switching as in your video from Noon Daylight vs Dual illuminant DNG profiles as seen in your blue shirt/orange pants kid. I do get color shifts between orange and blue colors switching between those two types of profiles.

Or do you have to use Lightroom 5 or versions of ACR newer than 4.6 and/or newer versions of DNG profiler to both build and apply the profile?

However, the ACR 4.6 defaults I build the custom profile, produce the most pronounced color shifts when switching between these two profiles than if I apply it to images that have Black and Contrast sliders set to zero and Medium Contrast curve set to Linear but there's still noticeable color shifts mainly in reds, oranges and some blues close to cobalt.

Another issue is WB adjusts from defaults profoundly change luminance pushing some saturated colors to clip and or bloom.

Your video makes the process seem far less problematic than what I've experienced shooting landscapes and local park scenery noon to late afternoon.

[digitaldog]

It confirms that illuminant specific based profiles don't affect colors the same way across all camera brands/models when switching as in your video from Noon Daylight vs Dual illuminant DNG profiles as seen in your blue shirt/orange pants kid. I do get color shifts between orange and blue colors switching between those two types of profiles.

I suspect too, the difference between 'canned' Adobe profiles and the one's we may build is based upon differences in the camera's consistency (our's versus theirs).

Or do you have to use Lightroom 5 or versions of ACR newer than 4.6 and/or newer versions of DNG profiler to both build and apply the profile?

Latest version of LR. All profiles built of course as shown in Passport software.

[Tim Lookingbill]

Latest version of LR. All profiles built of course as shown in Passport software.

Not clear what you're saying. I'll ask it in this way...

Have you seen an improvement in scene rendering from newer versions of the same software that both build and use custom DNG profiles and applied to default settings the profile was originally built from?

[digitaldog]

Have you seen an improvement in scene rendering from newer versions of the same software that both build and use custom DNG profiles and applied to default settings the profile was originally built from?

I'm not aware of newer versions of the Passport software. So I guess the answer is no.

[Tim Lookingbill]

Note: I'm aware that a persuasive case can be made that camera profiles tend to not be particularly worthwhile, even for reprographics where accuracy counts. "Do it by eyeball so it looks good."

Eyeballing it works most of the time especially when shooting scenes lit under various kinds of light. Check out Andrew's YouTube video on skin tone correcting with Lab vs CMYK. It demonstrates the a/b formula is a fairly predictable way to check skin color as long as the light is reasonably close to D50 like say in studio shots as in the Roman 16 image samples he uses.

Some sample images that even turns that on its head surprised me when I checked Lab numbers of Blu-Ray restoration screengrabs of Technicolor movie skin tones lit by what looks like extremely blue violet arc lamps. Most of the a/b numbers were equal indicating it should look slightly reddish but to the eyes looks slightly yellowish. This restoration in particular...

http://www.dvdbeaver.com/film/dvdreview/tocatchathief.htm

[JonathanRimmel]

I would say when using high iso >800, color changes. So unless you want to be really picky, I would only do separate profiles for iso's 800+. Perhaps a profile for 100-400, then one for 800, one for 1600.  It would depend on the camera obviously.

[xpatUSA]

If there are differences beyond some threshold, would it be worth the trouble to make different camera profiles for different ISO's?

Note: I'm aware that a persuasive case can be made that camera profiles tend to not be particularly worthwhile, even for reprographics where accuracy counts. "Do it by eyeball so it looks good."

According to Bogost, it would certainly be worth it for a Sigma DP1:

http://www.bogost.com/blog/technical_evolution_and_creati.shtml

Other Sigma models are said in the literature to vary with ISO (color shift as opposed to the usual noise whine).

For this question, how about accepting the premise that camera profiles are worthwhile, even for non-reprographics, to get a reasonable baseline starting point for subsequent p.p.?

I accept the premise but (OT alert) I'm given up fighting ACR 5.4 which only has one profile for my cameras and won't discover any new profiles I make no matter where put them on my HD. Therefore, RawTherapee is my serious editor of choice 

[Shutterbug2006]

I can only speak to the process of DNG profiles built with X-Rite's and Adobe's software, there's no reason to worry about ISO (or for that matter, building lots of DNG profiles for each scene). For odd illuminants (Fluorescent, LED) yes, you want to build a custom profile and be done.

This new 30 minute video on DNG camera profiles discusses this and shows the effects of ISO 100 vs. 800 upon the target (spoiler alert, there's none):

Nice video, and very well articulated. No doubt you help a lot of photographers grasp concepts that ordinarily are just beyond their reach.

Two thumbs up!

[D Fosse]

Maybe I'm stating the obvious, but just so we're all on the same page regarding color shifts.

A linear shift in color balance, if that is what happens with higher ISO, does not require a separate profile. It'll just be masked by the white balance. A new profile is only needed if the shift is non-linear and non-spectrally uniform.

I'm on a slow connection and haven't had a chance to check Andrew's video, but I'm sure it's covered there.

[mouse]

Andrew,
Many thanks for your video.

I have been able to create a dual illuminant profile with X-Rite software from shots of the Passport CC using my D700.

However using shots taken with my D800, the software cannot create a dual illuminant profile.  When attempting to load the second image I get an error message stating that the file is not a DNG image.  Of course this error message is erroneous.  I have tried contacting X-Rite about this problem, but no reply.

Anyone have a solution for the X-Rite software?  I have used Adobe's DNG Profile Editor, which is successful.

[digitaldog]

However using shots taken with my D800, the software cannot create a dual illuminant profile.  When attempting to load the second image I get an error message stating that the file is not a DNG image.  Of course this error message is erroneous.  I have tried contacting X-Rite about this problem, but no reply.

Windows? If so, someone else has reported the same issue, sounds like an X-rite bug. Did you try using Adobe's software instead?

[mouse]

Windows? If so, someone else has reported the same issue, sounds like an X-rite bug. Did you try using Adobe's software instead?

Yes, as noted in my post.  Adobe DNG Profile Editor works just fine.