Saturday, June 02, 2007

Photography is NOT reprography: Raw workflow needs tandem profiles.

Photography is not Reprography.

Dual ICC profiles as faithful and creative enablers for the Raw camera workflow.

I believe dual ICC profiles, working in tandem, are required for a creative Raw conversion workflow. The first profile is needed to define the colorimetric properties of the camera, the second to describe the operator's chosen exposure curve compression, color adjustments and other esthetic effects.

Photography is not reprography. Cameras are creative tools with which by pro photographers have historically used emulsions to create imagery; with respect to what an eye on location would see, the resulting individual images recorded on a chemical emulsion are often far from colorimetrically correct in any sense, even less so when lens filters or light gels are used creatively rather than for correction.

However existing digital ICC input profile workflows inherited from scanner days are historically designed to channel data with some degree of colorimetric correctness. To summarize, the scanned film workflow translates into accurately digitally reproducing the wildly inaccurate colors of an image chemically imprinted on the film.

In the digital playpen, photographers again expect to play with new digital versions of their old deliberately inaccurate creative tools. And the right moment for creative tone curve, color hue and exposure tuning is really during Raw conversion. Coincidentally, Raw conversion is also the moment where colorimetric accuracy is best established. Which is why we expect ICC profiles to find their use in dual ways during Raw conversion. And if the use is dual, two profiles should be applied.

Separating the Raw digital camera profile in two stages, one devoted to calibration —sensor and lighting characterisation— and one devoted to the "film look" would enable a more intuitive workflow. Essentially we argue that the creative steps can and should be follow the somewhat faithful ones, but both can be represented by ICC profiles.

The first profile, the calibration profile, is a standard ICC input profile which characterizes the camera and and Raw converter. The second which applies the look might be called a "digital emulsion". It can be realized by means of an abstract ICC profile. There is no need for the digital emulsion to be colorimetrically accurate in any sense - an emulsion may even emulate black and white film !

What software changes are needed for the tandem-profile workflow ? Clearly, new Raw conversion tools tools must allow both "calibration" and "look" profiles to be selected, and their superposed effect visualized. Present-day camera profiling tools are already adequate for the creation of the "calibration" profiles. Ideally however, a profile editor would allow the user to cook up "digital emulsions" on the fly so that they can then be applied to batches of images, even across camera brands.

Edmund Ronald.


Pix2Pixels said...

Obviously there are pro photographers and repro photographers.

Anonymous said...

Hello, Edmund, my name is Chris.

This is a little off topic from your post, but I came across your posts on photonet regarding adobe's approach to raw conversions similar to my experience.

Basically, I find their "profiles" to be complete junk, and thinking of all the time I (and others!) have wasted trying to get my color right because of these worthless things really irritates me. They are more of a half assed gesture in the direction of a profile than real profiles.

I have run the fors and rags scripts as well as calibrated manually with color samplers, none provides accurate calibration with ACR because ACR does not provide access to curve or even gamma adjustments to individual channels, and these are aparrently necessary.

I have also stumbled across a part of adobe's calibration "methodology", I think. If you run the rags script, and choose a2, b2, and c2 as the target patches, the script suggests 00,00,00 as camera calibration. The hues are still way off, even having given up on the saturation by targeting less saturated patches.


I conclude that ACR as is cannot be calibrated, and feel ripped off and insulted.

Thank you for telling it like it is.