7 Replies Latest reply on Jun 13, 2012 3:56 PM by sgtechsl

    Colorchecker Passport correction

    sgtechsl Level 1

      I know this has been discussed before, and probably should be a FAQ, but I've looked at previous discussions and have yet to find a definitive answer. Appreciate any tips.

       

      I think the problem is simple: I need to color-correct a RAW image that has a 24-patch Macbeth card in it. It's shot in full sunlight with a 45° angle between the lens and the sun. I know what the RGB values of the card are supposed to be:

       

      http://xritephoto.com/ph_product_overview.aspx?ID=1257&Action=Support&SupportID=5159&catid =28

       

      I need to do whatever needs to happen to this image to make it so that when I sample the various patches with the eyedropper in Photoshop, they're reasonably close (∆E of 7-8 would probably be adequate) to what the above chart says they should be.

       

      I looked at Rags Gardner's work:

       

      http://www.rags-int-inc.com/PhotoTechStuff/AcrCalibration/

       

      But it seems to have stopped at CS5/ACR5 and I can't get it working with CS6/ACR7.1 (so far).

       

      Is there a way to "pull" the colors in the image however they need to be shifted so that the 24 Macbeth card samples are reasonably close to what they should be according to the X-rite (or Lindbloom) charts say they should be, and the rest of the image's colors are calibrated along with it?

        • 1. Re: Colorchecker Passport correction
          Ron Parker

          If the chart takes up enough of the image you can use the Colorchecker Passport software or Adobe's DNG Profile Editor.

          http://xritephoto.com/ph_product_overview.aspx?ID=1257&Action=Support&SoftwareID=986

          http://labs.adobe.com/wiki/index.php/DNG_Profiles#Downloads_and_Installation

          • 2. Re: Colorchecker Passport correction
            sgtechsl Level 1

            Hi Ron, thanks for the reply. I hadn't tried the DNG profile editor before, so I gave it a whirl, but the results just aren't even close, so it feels like I'm missing a step or having some other problem.

             

            I opened my DNG image in the DNG profile editor and followed the tutorial for using the chart, positioning the four dots into their respective squares and creating a profile. It seemed to be successful. Then I opened the DNG image in Adobe Raw 7.1 and applied that profile, and opened it as a proRGB image. I would expect that now the colors should be correct, but when I sample (for instance) the blue square in the patch (#13 in this chart: http://www.rags-int-inc.com/PhotoTechStuff/AcrCalibration/D1X_2248_defalult.jpg ), that image says that the blue square's RGB values should be 26 / 57 /142, but when I sample that square with the eyedropper in PS I get 72 / 60 / 174, close on G but wildly off on R and B. RGB deltas in the 35-45 range can't possibly be within acceptable error ranges, I would think.

             

            What I'd LIKE is to have the magic profile adjustment tweak the image so that when I sample each of the squares in the chart, I get RGB values that are close to what the references for the chart say they should be. Close meaning maybe less than 10?

             

            Am I off somewhere in either my expectations, or my method?

             

            - Steve

            • 3. Re: Colorchecker Passport correction
              Noel Carboni Level 7

              What color space are you converting into?  That's going to make a difference in the RGB numbers.

               

              -Noel

              • 4. Re: Colorchecker Passport correction
                sgtechsl Level 1

                I've used ProPhoto so far, just for the widest possible gamut.

                • 5. Re: Colorchecker Passport correction
                  Noel Carboni Level 7

                  The test chart / process should be telling you what color space to measure the RGB numbers in.

                   

                  You don't get the same numbers in the different color spaces.  That's exactly the idea.

                   

                  Just as a guess, the given numbers in the image you put up probably refer to the sRGB color space.  Or maybe (the default Camera Raw conversion space) Adobe RGB 1998.

                   

                  -Noel

                  • 6. Re: Colorchecker Passport correction
                    ssprengel Adobe Community Professional & MVP

                    I think your expectations are way too high for how little color-error there should be using any sort of calibration process.  One thing to understand is that profiles optimize for the color of the light, not the luminance, and both things are tied up in the RGB numbers you are comparing.  Analyzing the difference in HSL or Lab and ignoring the L portion of either would be less misleading.

                     

                    I have not let any run to completion on my slow computer, but the calibration scripts seem to work ok in CS6 as long as you set PV2010 instead of PV2012 and perhaps reset any other parameters to reasonable zero/default values that have evolved since the scripts were last tested. 

                     

                    I prefer Tindemans’ script over Rags’ because Tindemans’ uses all the patches to compute the color error, not just the single primary color patch that Rags’ does.  Tindemans’ script also lets you set weights for the various patches so you can optimize for skin or landscapes or some other set of colors, because the whole set will never be all that close. 

                     

                    http://www.21stcenturyshoebox.com/tools/acrcalibrator/

                     

                    A few years ago I investigated various color profiles and used ColorCheck module of Imatest Studio (a few years back when it was a bit cheaper) to plot the color error:

                     

                    http://www.pbase.com/ssprengel/_tech_adobecalibration

                     

                    Others contributors on these forums have suggested that because the calibration scripts and the DNG Profile Editor both tweak different parts of the Adobe profile, that you could compute the slider values using the script, and then compute a DNG Profile from an Adobe profile after setting those slider values, instead of using the defaults of 0, to get a more accurate customization.  I have not attempted this, myself, but I understand the logic.

                     

                    The Adobe profiles have hue-twists in them that can produce more pleasing results for real photos than profiles created with the DNG Profile Editor. so while I used to use my script-based or DNGPE-based profiles almost exclusively, now I use the ones from Adobe unless I am dealing with a subject that looks good with stronger colors or I have a unique lighting situation that might benefit from a custom profile.  For example, recently I helped shoot a wedding in a church that had white walls and an unfortunate row of lighting along the sides, with varying shades of yellowish and greenish incandescent and fluorescent bulbs.  While, I did take some calibration shots before the ceremony to calibrate to the main overhead lighting, the custom profiles I created intensified the mixed-lighting color spots along the white walls—perhaps more photometrically accurate but ugly in this scenario.  Instead, I mostly used the Adobe Standard profile or sometimes the Camera Standard profile for the bulk of the shots.

                     

                    I do like using my custom profiles for outdoor scenes with bright colors in daylight and have some specialized profiles for sunset and twilight to take some of the edge off those extremes of ambient lighting.

                    • 7. Re: Colorchecker Passport correction
                      sgtechsl Level 1

                      ssprengel wrote:

                       

                      I think your expectations are way too high for how little color-error there should be using any sort of calibration process.  One thing to understand is that profiles optimize for the color of the light, not the luminance, and both things are tied up in the RGB numbers you are comparing.  Analyzing the difference in HSL or Lab and ignoring the L portion of either would be less misleading.

                       

                      That sounds right, and is probably pretty central to the misunderstanding I've had. It occurred to me that, of course, if I were attempting to calibrate two shots of the same scene, one in direct daylight and one on an overcast day, that it wouldn't make sense to expect the colorcheckers' squares to be identical after correction. One's in sunlight, so that one will be more luminant. Of course.

                       

                      So I guess the important thing to do is correct hue, and that can almost be done with just a white balance card, probably, at least for daylight. Photography under artificial light might need more help to address metameric issues and uneven spectral distributions, I would assume, but I guess the better process for correcting outdoor photography would be: either include a gray card or set a manual white balance in photography, adjust exposure settings until the image's histogram isn't clipped at either end, then take the photo and look at the resulting image on a properly-calibrated monitor and adjust for lightness until it looks "correct" to the eye, and declare it done. Does that sound more plausible?