Making profile calibration in Profile Editor with the GM ColorChecker I noticed that the original profile is always set to ColorChecker. I supposed there is anyway a original profile to which the calibration correction shift is applied to obtain the final profile to save. Is this "starting" profile the latest ACR 4.4 profile used in ACR or some other? In the PE FAQ page this is not detailed.
In fact I suppose the PE calibration is making only a "relative" adjustment related to a initial profile color characterization.
Thank You to anyone can give an answer.
MadManChan2000, Jan 8, 2009 6:46 AM
Massimo, yes, the base profile "ColorChecker" uses the color matrices that we build internally and thus are like "ACR 4.4" profiles. The reason the base profile is not "Adobe Standard beta 2" or one of the other new external (.dcp) profiles is because those contain lookup tables for performing output-referred effects (or "looks", if you will). These tables are replaced when using the PE's Chart Wizard feature.
The Chart Wizard is designed to produce a reasonably accurate scene-referred color profile (within the technical limits of the input data).
I need a little clarification here, please: from what You say I can safely assume the ColorChecker derived profile is not a output-referred profile (to give only some "looks") but a true "full new" profile that substitute all the ACR 4.4 data with new scene-referred measured values from the chart analysis?
Yes and no, Massimo. Let me explain. A profile contains several pieces. One piece is a set of color matrices. This handles a linear component of color correction (the bulk of what needs to be done). A second piece is a set of lookup tables which are used to handle non-linear color corrections (which cannot be handled by the first piece). When you build a color profile using the DNG PE's Chart Wizard feature, the resulting profile derives its first piece (i.e., the matrices) from the pre-built matrices that Adobe has already created from its lab measurements for your camera model, but it derives its second piece (i.e., the lookup tables) from your specific image of the ColorChecker.
You are correct in that the profile is NOT output-referred.
I need another detail, please: if the ColorChecker created profile uses a single 6500K CC image and a "6500 K only" setting, the resulting profile uses the new (measured from ColorChecker) lookup tables 5000-6500 daylight temperatures AND the OLD ACR 4.4 lookup tables for artificial light temperatures? So is a somewhat "mixed" profile that interpolates other temperatures using the new CC 6500K measure and the old ACR 4.4 measure?
Instead if I made a 6500K CC image profile with "Both color tables" setting, the resulting profile applies THE SAME lookup table for ALL the temperatures?
I think the obvious answer is yes to both points, but I would like to have a final word from You.
Can You please clarify if the Chard Wizard generated profiles are more or less influenced if the ACR sliders are moved from defaults?
The Camera Matching are set to have matching colors at the default settings of ACR. Because Chart Wizard profiles impact the same HSL tables that are used by Camera Matching profiles, can we have some impact on color fidelity if some sliders (brightness, contrast, exposure) are moved? If there are impacts, at which extents are visible?
> Can You please clarify if the Chard Wizard generated
> profiles are more or less influenced if the ACR sliders
> are moved from defaults?
Not exactly sure what you mean.
The Chart Wizard-generated profiles are intended to be fairly accurate from a scene colorimetric point of view, within the technical limits of the method. Any additional adjustments based in Camera Raw (i.e., beyond the defaults) are essentially "rendering" choices made by the photographer on top of the default color rendering provided by the profile.
> The Camera Matching are set to have matching colors at the
> default settings of ACR.
> Because Chart Wizard profiles impact the same HSL tables
> that are used by Camera Matching profiles, can we have
> some impact on color fidelity if some sliders (brightness,
> contrast, exposure) are moved? If there are impacts, at
> which extents are visible?
There are 2 sets of color tables in the new DNG 1.2 profiles. The main distinction is that one is 2.5D and takes place prior to exposure compensation, whereas the other is 3D and takes place after exposure compensation.
Profiles generated from the Chart Wizard only impact the former type of HSL tables (i.e., 2.5D) and do not contain the latter type of HSL table (3D). This is in keeping with the idea that Chart Wizard profiles are designed to provide fairly accurate scene-referred colorimetric estimates.
In contrast, Camera Matching profiles only used the latter type of HSL table (3D) but do not use the former type of HSL tables (2.5D). This is consistent with the idea that CM profiles are designed with specific output-referred effects in mind (e.g., lightness-dependent hue twists).
In general, the rendering controls in Camera Raw which affect tone and color take place __after__ the profile is applied, so you can think of these adjustments as being placed "on top of" the default rendering provided by the profile.
So the profile is a starting point and then you can adjust brightness, contrast, etc. relative to that baseline.
I didn't "hit" the DNG 1.2 specs about these two HSL lookup tables and their specific use.
A last question to confirm a little detail, please: You told that the Chart Wizard does not generate ANY 3D table. But if I manually modify some of the Chart Wizard generated colors in PE (i.e. I move the points on the color wheel after the GM panel has been read), the changes are written in the 2.5D HSL table anyway or they became a 3D correction?
Massimo, that would still be a 2.5D correction. The input to the correction is still 2D (it lies within the 2D color wheel). The output is 3D (H, S, and L adjustments). So the results still ultimately get baked into the 2.5D table.