Digital Light & Color

Just a background, so that it doesn't look like this is entirely academic:

I don't do studio photography, so I haven't had much interest in Match Reference. My jeweller friend asked me for help in setting up a workflow so that the photographs of his products look, within reason, somewhat similar to how products themselves look like when customers see them first hand. He's been struggling with it for ages.

Not knowing any better, I told him it should be simple enough: control lighting and camera settings, profile the camera, callibrate and then profile the monitor, and use PWP with colour-managed workflow - producing an output image conforming to and tagged with one of a standard ICC colour profiles (e.g. sRGB for web output, or printer-specific ICC for printing) so that what others see resembles what we saw. Simple...

With all else under control, we tried to use Match Reference, hoping that this could 'profile' the camera - i.e. produce a set of standard corrections that would reverse the distortions and quirks that camera/RAW-processing system introduces to the colours being photographed.

Some of the issues that we had are described above.

I'm experimenting to find out if building a Camera Profile instead of Matching Reference makes it easier.

A number of the patches in the ColorChecker are outside most monitor gamuts and are therefore cannot be reproduced onscreen. What it should look like is a simulated ColorChecker image in the monitor color space.

A better approach for things like catalog photography is creating a custom profile for the lighting setup, camera, and camera settings and then applying this to the images to correct them. Nothing will be perfect, but a typical IT8 calibration target has a lot more patches than a ColorChecker so the color correction can be a lot more precise than Match Reference.

Thanks Jonathan,

I'm learning how to do things with Camera Profiles - would you be kind enough to comment on what one should be aware of, as compared to Match Reference?

Would clipping grays outside of Macbeth 'white' and 'black' patches and massive colour shifts at different exposures be as much of an issue? More patches may improve overall correction precision, but those two problems with Match Reference were, I believe, independent of possible interpolation problems between sparse patches of Macbeth.

The Camera Profiling uses different method then the RGB Curves + Correction though? Could you briefly describe what it actually does?

Cheers!

The profile deals with blacks and white outside the range of the target, although a good digital camera target such as the ColorChecker SG has denser blacks than the regular ColorChecker. In any case, avoid clipping the blacks and whites in the target and try to expose so they are centered in the histogram. Try to avoid uneven lighting and also multiple lighting sources with different spectral characteristics - so all sunlight or all flash or all photofloods if possible. Also avoid lights with weird spectra such as fluorescent or LED.

Creating the profile involves finding the 3rd order polynomial in three variables (typically L,a,b) that best maps the colors in the reference image to their ideal values as specified in the reference file that characterizes the target. Three 1D lookup tables are also created from the target's gray patches to basically apply gamma curves to the RGB channels on input. The polynomial (which has 51 terms) is then used to create a 3D lookup table that the color management system interpolates to map colors from input RGB to its internal reference color space and from there to output RGB. In short the gamma curves are similar to Match Reference but the color correction is considerably more sophisticated. You can create a profile from the regular ColorChecker, but there are not enough patches to compute a 3rd order polynomial so a 2nd order polynomial is used instead. As a check of the accuracy of the polynomial fit, the errors for each patch are calculated and displayed.