Thanks for the C.M. forum!

I've been calibrating my cameras for a while now using ACR and several different methods, notably Fraser, Varis and the Fors script. I'm the type that likes to know whats going on so I prefer manual calibration (Fraser or Varis) in comparison to a script.

Fraser and Varis (author of "Skin") approach camera calibration differently; this has led to some head scratching.

The Fraser method first requires 'normalizing' the grayscale patches of the Color Checker Chart before tweaking the calibration sliders in ACR; that is, matching the R=G=B numbers from an ideal target to that of my real (measured) target. In my experience this usually means lessing the overall contrast of the real target by a significant amount thus altering the Red, Green and Blue patch RGB values from the cameras "native" state.

In comparison Varis prefers no or minimal adjustment of the contrast or saturation sliders (no normalization of the grayscale patches) before assigning calibration values.

Here's my question: So why would I want to normalize my grayscale patches before assigning camera calibration values in ACR then turn around and apply these calibration values to images that have not been grayscale normalized? It seams to me, at least in "theory" anyway, that this method could potentially produce funky calibration values to real images because the calibration was determined in a significantly different environment than the average RAW image.

After much consideration I'm leaning towards Varis. It seems to just make more sense to calibrate to ideal Red, Green and Blue patch values based on a cameras native gamma rather than assigning an ideal gamma then calibrating. After all, aren't we just trying to modify the output our camera's sensor to that of "actual" (1st person perceived) color?

My guess is that I'm missing something. I've not read all of Bruce Fraser's work but what I've read is extremely impressive. For me to suggest that his camera calibration method seems theoretically misguided would be naive, but as I've thought things out intuition tells me that this approach may lead to less than ideal calibration values for Real World images (pun intended).


Explanation and thoughts welcome. Eric

What I've found is that with the Fraser/Fors approach you end up with a 'normalized' contrast setting that's very low, often zero or slightly negative. This in turn affects the saturation of the image, and as a result you end up with a calibration that has the saturation sliders on the calibration tab set quite high to compensate for the desaturation caused by the low conrast setting.

When you then use your calibration with real-world images where you don't have the contrast slider set to zero, the image is over-saturated. For this reason I use one of the alternative calibration scripts that allows you to skip the normalization of the grayscale patches and leave the brightness/contrast sliders alone. To me this gives better resuts.

Jeff,

The high saturation that one gets with the Fraser/Fors approach has bothered me too. What alternate script do you use?

Bill

I may be off, but my understanding is, that the normalization is necessary in order to compare the result of the calibration to the nominal values of the checker card. If you change the overall lightness, all the composition changes, except for those colors, which have all three components identical, or one (or two) components null and the other two equal. So, if the lightnesses of the gray patches do not match with the nominal values of those patches, then the colors of the other patches *must not* match with the nominal values - and what can you compare?

The result of calibration applies to all images, independently of their color composition.

The issue with using the Brightness & Contrast sliders before setting the Calibration sliders is that the CR tone curve is set to lower contrast. CR's tone curve preserves hue but not saturation (by design); hence lighter colors mapped through the tone curve become less saturated and conversely darker colors mapped through the same tone curve become more saturated.

To match the ColorChecker Chart's gray patches you typically a lower-contrast tone curve than CR's default tone curve. If you then set the calibration sliders to optimize color matching to the chart reference values (using whatever metric / script you prefer), then go back to using CR's standard tone curve, then the darker colors -- in particular the first three patches in the third row of the CC -- will become noticeably more saturated than the reference values.

If you're looking for the same color response as the ColorChecker then you need to keep that lower-contrast tone curve (determined by your adjusted Brightness, Contrast, etc. settings) both for the calibration step as well as when processing your real images.

(In practice the majority of photographers don't like using that lower-contrast tone curve because subjectively the images don't look nice.)

Determining calibration settings with a strong tone curve defeats the entire purpose of calibration--matching the captured colors to the original subject as exactly as possible. The most accurate capture must use a linear tone curve (gamma excepted). S-curves are a departure from accuracy applied for aesthetic/creative reasons. If you apply an S-curve and you think it increases saturation too much, keep in mind that ACR does have a saturation adjustment that allows you to reduce saturation if desired. It's really not a problem.

Sorry, lost track of this thread. Both of the scripts below have quite a bit more options than the original, including the option to leave the settings on the 'Adjust' tab alone:

http://21stcenturyshoebox.com/tools/ACRcalibrator.html
http://www.rags-int-inc.com/PhotoTechStuff/ColorCalibration/

A follow up to my original post and recent findings...


Calibration workflow: Make a target similar Varis as explained in his book 'Skin'; meter gray card and bracket shots up to 3 stops over spot meter reading. Import into ACR, (not Lightroom due lack of color samplers) zero then determine proper exposure of Row 4 Colum 3 patch (145 PPRGB). Set white balance using Row 4 Column 2.

I'm finding that normalizing the grayscale patches is in some respects camera specific.

Case in point, a properly exposed Row 4 Colum 3 patch on my Fuji S5 was 1 1/3 stops over the gray card reading. Based on this exposure all the Row 4 patches came out nearly identical to to the "ideal" ProPhoto target. No normalization was required and strait to calibration I went.

For my D300 the proper exposure for the Row 4 Colum 3 patch was two full stops over the gray card reading; because of the constricted dynamic range of the D300 it was necessary to "normalize" row 4, but each tweak was very minor and didn't leave me wondering about noticeable saturation shifts.

I guess my point in all of this is to spot meter a gray card (Robin Myers in my case) and then determine the proper exposure for Row 4 Column 3 patch based on the ProPhoto 145,145,145 ideal value. Using this approach seems to remedy much of my concern I expressed when using the Fraser method outlined in Real World Camera Raw.

Cheers, Eric