15 Replies Latest reply on Jan 23, 2009 10:16 AM by MadManChan2000

    Some questions about DNG PE

    Marco N. Level 1
      With a good calibration in 5000K lightning condition I have tweaked my 4.4 profile to match the ideal color value (read by spectrophotometer) of my CC so that the final average error is near DE2000 one.

      1. Is a good idea report this calibration setting in Color Matrices pane *before* create the two color table with chart wizard?

      2. Indeed my intent is to leave to profile the first management of the tone mapping like in the early version of ACR, because I don't understand the new Adobe approch. Why tweak only hue and saturation?

      3. Is it different starting from the 4.4 profile or from AS profile in DNG PE?

      4. What tollerance I have in lightning condition for the targets capture? For example, I shot the targets in 3200K and 6000K condition, is this a problem for the final result?

      Thank you and sorry for my English
      Marco
        • 1. Re: Some questions about DNG PE
          MadManChan2000 Adobe Employee
          > With a good calibration in 5000K lightning condition I
          > have tweaked my 4.4 profile to match the ideal color value
          > (read by spectrophotometer) of my CC so that the final
          > average error is near DE2000 one.

          > 1. Is a good idea report this calibration setting in Color
          > Matrices pane *before* create the two color table with
          > chart wizard?

          Yes.

          > 2. Indeed my intent is to leave to profile the first
          > management of the tone mapping like in the early version
          > of ACR, because I don't understand the new Adobe approch.
          > Why tweak only hue and saturation?

          You can tweak hue, saturation, and value (lightness). However, the lookup tables are only indexed by hue and saturation. This makes the LUT essentially 2.5D (indexed by 2 dimensions, and each table entry has 3 components).

          The new approach explained in DNG 1.2 (and implemented starting in Camera Raw 4.5 and Lightroom 2.2) is a proper superset of the earlier approach. That is, everything you could do before can also be done with the new approach; plus there are things you can do with the new approach (such as saturation-dependent hue twists) that could not be done with the old approach.

          > 3. Is it different starting from the 4.4 profile or from
          > AS profile in DNG PE?

          When defining your own color adjustments manually, yes, it matters which base profile you start from.

          When you are using the Chart Wizard, it does not matter which base profile you start from.

          > 4. What tollerance I have in lightning condition for the
          > targets capture? For example, I shot the targets in 3200K
          > and 6000K condition, is this a problem for the final
          > result?

          I would not worry about the 6000 K end; with 3200 K you may notice a bit more discrepancy since small changes in color temperature at the low end of the scale can account for larger color shifts.
          • 2. Re: Some questions about DNG PE
            Marco N. Level 1
            > 1. Is a good idea report this calibration setting in Color Matrices pane *before* create the two color table with chart wizard?

            >Yes.

            Is that pane indipendent for each illuminants or common?

            Thank you.
            Marco
            • 4. Re: Some questions about DNG PE
              Peter DL Level 1
              > 1. Is a good idea report this calibration setting in Color Matrices pane *before* create the two color table with chart wizard?

              > Yes.

              So, let me ask, why doesn't the chart wizard calibrate the matrix primaries first ?

              Thanks for your comment.
              Peter
              • 5. Re: Some questions about DNG PE
                MadManChan2000 Adobe Employee
                I considered doing that, and it may still be a good idea. Maybe in a future update. (In general, the more work handled by the matrix, the less work needed by the LUTs.)
                • 6. Re: Some questions about DNG PE
                  Marco N. Level 1
                  That is my starting idea, first a big global tweak via old calibration way and than small zonal tweaks for the last fine profile refinement, via DNG PE chart wizard.
                  In this first effort the Tindemans calibration script is actually an invaluable tool, driving masterfully the profile to match what the spectrophotometer tells about my CC (I calibrate in my most lightning shooting condition (D50)).

                  In this way I hope sooner :-) DNG PE could import the *spectral* value of our CC (in GMB/Xrite format), so that it can compute the real color value via CA starting from the WB lightning condition indicator (not 100% credible, but ...). In this way I think DNG PE could acquire flexibility (more tolerance in lightning condition when the target is captured) & accuracy.

                  Again as Bruce told "We do suggest that you measure the 24-patch ColorChecker yourself. The formulation has changed at least three times over the years and while the patches from each formulation appear similar to the human eye, they appear rather differently to digital cameras."

                  I add that (now I don't have with me my CC measurements) they are not so similar along the sample and format and they changes in the time.

                  Marco
                  • 7. Re: Some questions about DNG PE
                    Peter DL Level 1
                    One easy option and procedure is to feed the Matrix pane with the results from the Chart wizard, taking the readings for the three primary patches i.e. Blue #13, Green # 14 and Red #15. Then, clear all color adjustments and run the Chart wizard again. After 2 or 3 iteration of this kind (needed for fine tuning because the numerical scale does not directly match), the remaining work for the Lut is about zero as far as the three primary patches are concerned.

                    But then, Im wondering if the canned baseline Matrix was really meant to provide somewhat accurate results (within its technical limits, and subject to possible device-to-device deviations), or, if the canned baseline Matrix was already tweaked by design in any way to provide a preferred rendition? Just a thought.

                    Eric, any word on this?
                    Thanks again,
                    Peter
                    • 8. Re: Some questions about DNG PE
                      MadManChan2000 Adobe Employee
                      No, we did not do any additional tweaking to the matrix profiles.
                      • 9. Re: Some questions about DNG PE
                        Peter DL Level 1
                        Thank you, Eric, for the clear response!
                        In consideration thereof, let me finally ask:

                        Wouldnt it be better to (pre-)calibrate the matrix primaries as well when using the Color Matching profiles such as e.g. Camera Standard? Thus, to establish the 'same' starting point as for which their Lut was built i.e. set on the top?

                        If yes, what in detail needs to be done to be on the same page so to speak: should such (pre-)calibration of the matrix primaries target to get just the three primary patches right (#13-15 of the CC), or should we go for a best fit matrix across all color patches? Further, in the course of shaping the matrix primaries, should the ACR default tonal settings be left in place (Blacks 5, Contrast 25, etc.) so that their side effect on saturation is considered and compensated for, at least on a somewhat average basis?

                        Thanks again for your attention.
                        Peter

                        --
                        • 10. Re: Some questions about DNG PE
                          Tim Lookingbill Level 1
                          I can tell you from my experience calibrating to the CC chart, Bruce's suggestion of having to measure your own chart with a spectro may be in order because different lighting situations can give different results for each patch that makes the numbers that are included with this chart visually worthless.

                          One problem with calibration is that it works great in the studio where lighting, camera angle and exposure can be tightly controlled. Outdoor lit scenes pretty much makes going by the numbers even more worthless.

                          Here's the CC chart shot in direct sunlight using the Tindeman script pretty much nailing the Lab numbers that came with my chart. The red and yellow looks off, but my memory of this chart viewed outside doesn't quite jibe with the results.

                          http://www.pixentral.com/show.php?picture=1JUNQWIVRZI4Q0ro0XcPrfD0hCAPqt1

                          Here's two shots of the CC chart lit by an Ottlight CFL and a Solux halogen. The Tindeman script was useless and I had to resort to editing by eye which I have to say even ACR 3.7's calibration sliders are the best in nailing the color out of all five raw converters I've used trying to get this chart to appear as my eyes see it.

                          http://www.pixentral.com/show.php?picture=1WUXP8ixKrRQi6x6g1FFQTDGsFdj

                          The problem is using lights like outdoor sunlight to light the chart that requires the use of memory when editing which isn't very reliable especially shooting landscapes. Studio shots I'm sure the PE Chart Wizard and scripts should do just fine.
                          • 11. Re: Some questions about DNG PE
                            MadManChan2000 Adobe Employee
                            The reason the Chart Wizard doesn't adjust the matrices (e.g., a best fit according to some favorite metric across the 18 color patches) is because internally we already do a fit across a much larger set of patches under very controlled lighting. This tends to produce a more consistent starting point from camera to camera.

                            (For some cameras, unit-to-unit variation is automatically compensated for using calibration data that is already embedded in the raw file. Thus it's not useful to adjust the color and forward matrices (see the DNG spec for details) to optimize for the Camera Standard or other "Camera"-prefix profiles that we build.)
                            • 12. Re: Some questions about DNG PE
                              Peter DL Level 1
                              Eric, - I think I can follow your reasoning with regard to the AS and CM profiles. Provided that device-to-device deviations with the camera play a subordinate role, the idea to calibrate the matrix primaries first is probably prone to make things worse than better. Talking about possible unit-to-unit deviations with the ColorChecker, maybe an amateurish lightening and the limited number of the color patches for averaging as opposed to your pro environment.

                              However, such pre-calibration still seems to me a good idea in conjunction with the Chart wizard. While above mentioned sources of error may still apply, the procedure of shaping the primaries first results in less work for the Lut (just as you said) and leads to a more homogenous color adjustment as far as I can tell.

                              FWIW, the resulting profile is pretty much to my liking, even and in particular when judging in terms of 'pleasing' rather than 'accuracy'.

                              Many thanks for your patience.

                              Peter

                              --
                              • 13. Re: Some questions about DNG PE
                                MadManChan2000 Adobe Employee
                                Yeah maybe one option is to have a checkbox in the Chart Wizard as an option for whether or not to do a matrix calibration before doing the lut adjustment.
                                • 14. Re: Some questions about DNG PE
                                  Marco N. Level 1
                                  >No, we did not do any additional tweaking to the matrix profiles.

                                  I don't understand one thing: if you didn't do any additional tweaking, why the calibration is always a tradeoff along all patches of CC? For example, why two patch of the red domain like red and light skin have opposite hue tweaking necessity after the calibration?

                                  Marco
                                  • 15. Re: Some questions about DNG PE
                                    MadManChan2000 Adobe Employee
                                    Because they have different lightness values. Colorimetrically there is no need to distinguish between these, but in practice (i.e., when trying to model scene appearance, or for reasons of preference) you may wish to treat dark reds differently than light reds.

                                    The DNG PE does not have this editing capability (i.e., it only has a color wheel in 2D, not a full 3D editing ability which would require a more sophisticated interface).