1 person found this helpful
An interesting question, and I'd like to hear some of the answer, but I think you're in the wrong forum.
You might want to try the DNG forum too:
Thanks. I somehow missed that there was a DNG forum, so I've re-posted there.
White balance defines the neutral point, i.e., which recorded color is going to appear neutral (achromatic, gray, R = G = B, if you prefer). It does not say anything about how the non-neutral controls will appear. Will a given non-neutral color appear more yellow or more orange? Less or more saturated? These other characteristics are determined by the color profile.
At a high level, the profile is responsible for determining the mapping between raw colors (as recorded by the lens and sensor) and a standard colorimetrically-defined space (e.g., ProPhoto RGB). Because different cameras use different sensors, these mappings are usually highly specific to the sensor.
In any case: a profile created for one lighting condition can often work well for other lighting conditions, as long as the two lighting conditions are reasonably close. Performing a white balance is effectively a "normalization" that factors out the differences.
I think my lack of understanding is perhaps even more basic, and my guess is that many who are new to this have the same problem. I know the DNG camera profiles I've made (w/ColorChecker Passport) work great, I just can't figure out how they work. For example, I can create a single illuminant profile shooting a chart under tungsten light, but since I don't tell the system what the color temperature (or spectral distribution) of the light source was, how can the software use this image to adjust for camera/lens variations? How does the software determine which variations are due to lighting which are caused by camera/lens/sensor variations? A very basic question, yes?
Does the profile-building software first determine the color temperature by sampling the color-neutral samples and then normalize and calculate variations from what's expected in the color samples? But even so, how does the profiler distinguish between a local hardware variation and perhaps uneven distribution of the source light? For example, suppose the color temperature is 5000 K but there's a significant notch in the red range? How does the system know the difference between a lack of red in the illumination vs. a lack of red in the sensor? I don't know the proper terms, but it seems to me that color temperature only gives you some average for the distribution of color and doesn't tell you how evenly the light is distributed across the spectrum.
Hey, I know it works great. My profiles have saved me a lot of time and agony. I just can't figure out how my profiles that are shot under unknown lighting conditions can work for shots taken in others. I'm obviously just unclear on the concept.
FYI, I shot all my profiles using two exposures: one with a raw sb-900 strobe and the other with a TN-A1 filter over the strobe (to simulate tungsten). I haven't yet tried to use these dual-illuminant profiles to adjust images shot under conditions far outside that range such as cold fluorescents.
See this Luminance Landscape thread about dual illuminant profiles:
Scroll down to the Color Checker chart samples I posted to see the visual differences.
Your camera may deliver different results.
As to your other questions about what's going on under the hood with regards to color temp influenced color rendering, there are way too many variables to consider to offer a definite answer to improve workflow.
I've profiled my Pentax K100D but I still can't make the camera render the way my eyes see a certain piece of polyfiber yarn cloth lit under full spectrum D50 light. This cloth reflects a odd hue of grayish dull brown with a bit of yellow (raw umber?) that comes across in post as very reddish. Even selective color edits in Photoshop can't emulate the hue without introducing color artifacts.
Another odd optical effect color temp perception and manipulation has on the eyes is demonstrated below of a shot taken with my Pentax showing a dull, flat scene with a subtle cyan-ish blue cast. I adjusted the color temp in ACR to look as it does for demonstration purposes. Note the color cast seems to disappear just bumping up the saturation on the same image.
> I just can't figure out how they work. For example, I can
Adobe ships separate color profiles for each camera model. Each profile is specific to the sensor that was measured by Adobe. There could still be unit-to-unit variations. These are generally not addressed by the profiles Adobe ships (though in some cases they can be, see the Calibration tags in the DNG spec if interested). Generally, profiles also don't deal with differences in lenses. Some lenses are slightly warmer/cooler than others. Much of this can be factored out by white balance (e.g., photographing a neutral card and then click-WB on it), but not entirely.
> How does the software determine which variations are due
> to lighting which are caused by camera/lens/sensor
> Does the profile-building software first determine the
> color temperature by sampling the color-neutral samples
> and then normalize and calculate variations from what's
> expected in the color samples?
> But even so, how does the profiler distinguish between a
> local hardware variation and perhaps uneven distribution
> of the source light?
> For example, suppose the color temperature is 5000 K but
> there's a significant notch in the red range? How does the
> system know the difference between a lack of red in the
> illumination vs. a lack of red in the sensor?
> I don't know the proper terms, but it seems to me that
> color temperature only gives you some average for the
> distribution of color and doesn't tell you how evenly the
> light is distributed across the spectrum.
> Hey, I know it works great. My profiles have saved me a
> lot of time and agony. I just can't figure out how my
> profiles that are shot under unknown lighting conditions
> can work for shots taken in others. I'm obviously just
> unclear on the concept.
> Generally, profiles also don't deal with differences in lenses. Some lenses are slightly warmer/cooler than others.
Some time ago I made a study about the influence of five lenses on the profile in a digital system: if you want read the article click on this link http://www.photoactivity.com/Pagine/Articoli/052ResaCromaticaObiettivi/Resa_cromatica_degl i_obiettivi_ENG.asp