1 Reply Latest reply on Apr 8, 2006 12:43 PM by Peter DL

    Sense & Simplicity of ACR Calibration

    Peter DL Level 1
      During another recent discussion I mentioned the Sense & Simplicity of ACR Calibration and while no one explicitly asked, I thought I should explain what I mean. In the worst case this wastes a little bit forum space, and in the best case it could save someone a year or so which actually was the length of my learning curve to reach this conclusion: ACR calibration in fact makes sense and it can be quite simple.

      Let me start by emphasizing an obvious detail: The Calibrate Tab sliders actually do what the labels suggest; they adjust the red, green and blue Hue & Saturation with a main weighting on these primary colors and a decreasing influence on secondary colors, respectively. While this is computed behind the scenes on a CIE xy basis, there are enough similarities with HSB to trace the changes via the HSB readings in Photoshop (after Raw conversion). For example, the Red-hue-slider mainly changes the HSB-hue of red colors, though there are some second order effects on HSB-saturation & brightness.
      Anyway, the Calibrate Tab does not include any Brightness sliders for red, green and blue which would allow to compensate for relative brightness differences. The brightness of a color is as it is - which is mainly determined by the tonal controls. Given that we just can calibrate the Hue & Saturation, I suggest to desist from any attempt to linearize the grayscale and to eliminate the nice tone curve of Camera Raw.

      My eye-opener was a portrait of a person who was wearing a red shirt. The image was reasonably white-balanced based on a previous shot of a graycard, and the tonal controls were set according to taste. As always, skin tones were too magenta so I dragged the Red-hue-slider to the right (in Camera Raw), until I noticed that the red shirt was getting too green. Finally, an acceptable compromise was found; for the skin tones as well as for the red shirt.
      Further, there was some bright green grass in the background which was clearly to green so I dragged the Green-hue-slider to the left to add a touch of red.

      This Hue calibration can also be easily accomplished by means of a captured ColorChecker chart:
      /> Camera Raw initially set to the 2.x shipping defaults (PS CS); means 3.x without auto-corrections, and the Curve Tab set to linear (PS CS2). Some more details: Shadows 5, Brightness 50, Contrast 25, Saturation 0 (should be valid for most cameras), Sharpening 0, Smoothing depending on camera needs and all Calibrate Tab sliders at 0. Further, ACR set ProPhoto RGB at 16 bit precision.

      /> Whitebal the second gray, patch #20 by means of the eyedropper. Adjust Exposure to RGB= 190 for this patch.

      /> Make a first Raw conversion in order to compare the HSB-hue readings with the reference numbers as follows. The numbers were derived by averaging the simulated charts as offered by ColorRemedies as well by BabelColor:
      red patch # 15: H= 10°, S= 62%
      # 2 light skin: H= 30°, S= 28%
      green patch #14: H= 101°, S= 45%
      # 11 yellow green: H= 75°, S= 56%
      blue patch # 13: H= 248°, S= 60%
      # 3 blue sky: H= 230°, S= 29%

      /> Open the Raw file of the captured chart again and adjust the Hue-sliders of the Calibrate Tab in an intuitive way to eliminate the numerical difference compared to above target numbers. Convert and compare again.

      /> Everything should be done after two or three rounds and Raw conversions. The main challenge is to consider two patches simultaneously, e.g. #15 red and #2 light skin while adjusting the red hue, thus, to reach a best possible compromise (if necessary).

      Now what about saturation? It is no problem at all to add some further rounds of Raw conversions in order to adjust saturation. As far as I can tell, this almost does not distort the hue again. The question is just, if a calibration of saturation makes sense.
      In my case, the ColorChecker wants me to increase the green saturation. This is definitively the wrong way to go for the image I described above
        • 1. Re: Sense & Simplicity of ACR Calibration
          Peter DL Level 1
          Oops ...

          In my case, the ColorChecker wants me to increase the green saturation. This is definitively the wrong way to go for the image I described above. The green grass in the background was quite glooming anyway; to increase its saturation would have been completely contradictive for a pleasing rendition. And, theres probably a very good explanation for this finding.

          By getting color saturation perfectly right, the de facto tonecurve as set forth by Shadows, Brightness and Contrast is silently turned to a HSB brightness curve. And here we meet the known downside of the HSB color model: different hues are numerically treated the same. Said de facto tone curve which is an elegant brightening S-curve finally leads to a situation where cyan-green-yellow hues are gaining too much visual dominance and Lab-lightness.

          Human eyes (at least mine) call for a certain hue-accuracy, but the idea of an accurate saturation lets us run into the difference between an accurate vs a pleasing rendition.

          Theres certainly more than one way to deal with this situation. For my purposes I found it useful to implement a calibration with accurate red-saturation and slightly under-saturated green. But this is just a starting point to tweak the Saturation-sliders of the Calibrate Tab depending on image-specific needs.

          Some final remarks:

          I found that the procedure as suggested above leads to quite reproducible results comparing different captures of the ColorChecker target. All shots were done at full spectrum lighting via direct sunlight which also should minimize metamerism. However, I cannot comment on the influence of in-door light sources and studio conditions. Further, I could not find any relevant influence of different exposures +/- EV.

          Whereas the main tonal controls: Shadows, Brightness and Contrast have a significant and complex side effect on color saturation like we know it with RGB curves. Certainly an interesting chapter from a math perspective, but sometimes not without problems in practice. Inevitably, the settings for Shadows, Brightness and Contrast are part of the calibrated state. Changing away from it, more or less changes color saturation. One trick to reduce such effects is to maintain the ratio of Brightness : Contrast = 50 : 25 = 40 : 20 = 30 : 15, etc. if this is acceptable depending on image content.

          Like with any ACR calibration procedure, the final success depends on the Bayer R/G/B filters and its spectral response. In the ideal case all colors would snap into place upon getting just the three primary patches #13-15 and the three CIE xy primaries right. With real-world cameras its probably better to consider more color patches for a best possible average. Above suggested memory colors: light skin, yellow-green and blue sky of course just reflect my personal preferences.

          With best regards, Peter

          Some References: