13 Replies Latest reply on May 27, 2011 11:42 AM by Hudechrome

    CCP colour profiles and different lenses

    Mollysnoot Level 3

      Hi,

       

      I just got a Nikon D7000 and I've been playing around with my ColourChecker Passport to set up some standard colour profiles for use in ACR as a general starting point for processing. I've been pondering if it's worth my while to create different profiles for each lens I have, something I've not previously done when profiling my old D60, where I just created a series of profiles (including some dual-illuminants) by using one lens and capturing the target under a variety of different lighting conditions (e.g. tungsten, flash, sunshine, etc).

       

      Anyway, I just tried creating a profile for my 105mm 2.8 lens under tungsten lighting, having previously (yesterday) created one under the same lighting with my 50mm 1.4 lens and I've been comparing them in ACR using the colour dropper. I’ve opened up the images used to create the profiles, applied the profile generated using the ColourChecker software for the corresponding lens, and then set the white balance using the ‘off-white’ colour patch with the eye dropper WB tool. I then used the colour dropper on the same colour patches in each image. I’ve noticed that the RGB colour values aren’t matching quite as well as I’d expected (note that I thought it potentially unrealistic to get a perfect match): blues and greens seem to be roughly the same, so for example with patch #3 (third from left on the top row), one is at 69,72,115 and one at 70,77,115, but reds and oranges seem to be a bit further out of sync, e.g. with patch #15, one is at 99,45,29 and one at 109,51,34; with patch #16 one is at 166,167,29 and one at 175,179,33. This surprises me a little, as I thought the idea of CC was to calibrate the profiles so that colours were essentially the same across different lenses – and different cameras if applicable. I have to say though that, colour values aside, when eyeballing the two images on my monitor (profiled) they do look very similar, which I guess is the main thing!

       

      I wonder if perhaps I’m missing something here? I’m quite prepared to be told that I’ve got this all wrong!

       

      Also, I wonder if others on the forum using CCP have gone to the trouble of creating lens-specific profiles, or if they’ve just created profiles for their camera body using one lens? This is the approach I took with my D60, but having done more reading on CCP I know that some folk do advise to create separate profiles for each lens they use (and I am of course aware that the CCP user manual also states to do this). Do you even create a profile for each and every shoot (when possible)?

       

      I’d be very interested to hear your opinions on this as I’ve not been using CCP for all that long and am always eager to learn more.

      M

        • 1. Re: CCP colour profiles and different lenses
          Jeff Schewe Level 5

          In the grand scheme, you don't need to do different profiles for similar lighting such as flash and daylight. The differences will be negligible. You do want to do dual illuminate profiles if you can. Somewhere near D65 (or D55) and tungsten. This will help in the accuracy of the "tweening" between lower and higher color temps. If you have a specialty lighting that is non-standard such as florescent, it IS best to do a specific profile for that lighting...

           

          As far as profiling for lenses, unless you have a lens that has a clear and repeatable lens cast, I wouldn't bother with lens specific DNG profiles. It would be useful to be able to create lens cast corrections. Some lenses are warmer/cooler but at this point it's not really possible with ACR. Hopefully that may change in the future.

           

          I suspect that some of the reading differences you may be seeing may actually be tone and contrast differences between the lenses you tested. It's real hard to nail the exact exposure between shots when you change lenses...there are variations in F stops between different focal lengths. Also note that tungsten lighting is a real challenge for really accurate color because the blue channel is so under exposed compared to the red and green photo sites. It's the nature of the spectral response of the color filters on the sensor.

          • 2. Re: CCP colour profiles and different lenses
            ssprengel Adobe Community Professional & MVP

            First of all, a color profile is for correcting color, not luminance, so compare the HSL or Lab coordinates not the RGB values so you can just ignore the L coordinate.  From your given RGB numbers, you can already tell that one of the images is brighter than the other so it is just confusing looking at the RGB values and guessing what you would expect the three values to be in the other image.  For comparing two images, I would concentrate on the Hue number in HSL coordinates, since Saturation can change with contrast, and Luminance can change with Exposure and Contrast.

             

            Also, as part of your eyedroppering comparison, another thing to do would be adjust the "Exposure" of the darker image until the L number (in HSL or Lab) is the same as the L in the brighter image and then see what the other two numbers are--maybe the other two numbers won't change, and then you can try putting one of the HSL values in the "Old" patch of the color-picker and the other in the "New" patch and see how much different they look.  You'll have to do this comparison in Photoshop not ACR so use ProPhotoRGB when you export to keep the colors as close to the same as you can.

             

            The two questions you seem to have, are:  does using a lens-specific profile make enough difference to real world situations to bother with, and where are the variations I'm seeing when the profiles are applied to their source images coming from since I would think they would be the same.

             

            For testing whether the profiles computed for the two lenses make a noticeable difference even with your two profiles that don't appear to correct the same, apply the two profiles to the SAME CC image (one of the two you created your profiles with), save an sRGB JPG of each, and see if you can tell the difference, either side-by-side, or even better, when you flip back and forth in some sort of photo viewer--like with Windows Picture Viewer when those are the only two images in the folder.  By apply the two profiles to the same image you have mitigated any luminance and white-balance differences in the source image and are merely looking for differences in the effect of the two profiles. 

             

            If you can't tell much difference between the same image using each of the two profiles then it's just an academic exercise.  I like academic exercises, but am also a perfectionist and lazy so I would do the experimenting until I found out I'd perfected things enough that I can't tell any difference then I can stop.  In other words, do I need to profile for various lenses or not, or am I just doing it because I like to control everything as much as possible and it really doesn't make any difference. 

             

             

            Before answering the other question, about where any profile variations might be coming from, understand that the combination of white-balance and color-profile is attempting to convert the colors of an object photographed in the lighting scenario the profile was created for into the colors of the object photographed in a standard lighting scenario.  In my mind the works out to be "make the colors of the object look like it was photographed in sunlight".  The issue that requires making a profile and not just white-balancing, is that any part of the object that was colored the same as the light color will be neutral when the white-balance is done, and more generally the closer the color of the object is to the color of the light, the more neutral it will become when WB is done.  For example, if you have a red ball and a gray ball and photograph them in red light, they will both look gray when white-balanced.  A real-world example of this would be flesh-tones in incandescent light, when white-balanced will have even less color and be more neutral or pale or even bluish, than the skin photographed in sunlight, so after white-balancing, the job of an incandescent profile is to boost the reddish colors and diminish the bluish colors so the skin looks like it would in sunlight.  This might be an argument for NOT WBing skin in incandescent lighting.  In severely-colored lighting, especially nearly monochromatic lighting such as sodium vapor lighting, correcting the colors to be as if in sunlight will be impossible, but to the extent the lighting isn't monochromatic, the colors can be made to look more normal, if not perfectly normal..

             

             

            To understand whether the differences you're seeing in the profiles are due to the lenses being different color or due to variations in the profiling process, itself, think about where the variations could come from and how you might test for each: 

             

            Was the source lighting exactly the same color between the two shots with different lenses (that were taken a day apart)?  Test by eyedroppering the WB of same neutral-color patch in each photo and see if there is any difference in the Temp/Tint numbers.  You cannot test the source-lighting color unless you have shot with the SAME lens for both days, so if you don't have shots with the same lens, seeing that the WB is not much different between the two shots can give you some comfort that the difference in the profile was not a difference in the source lighting.  The source lighting might have changed if there was some daylight mixing in on one day and not the next, or if the A/C was running on one day and not the other and the voltage was slightly different and the redness of the light was different.  One other thing that can wreak havoc in repeatability of both color and exposure is if any of the lighting is fluorescent CFL or tubes, because that sort of gas lighting changes intensity as the voltage varies and reverses 60-times per second and this variation is especially noticeable if the shutter is fast.  So while your lighting may have been incandescent any changing daylight or flickering fluorescent lighting mixed in might have changed the source-lighting color enough to make a variation in the profile more than the color of the lenses might have.

             

            This first question dealt with the photos taken with each of the two lenses.  The remaining questions are about testing with just one lens. 

             

            Is the profiling process repeatable?  Test by creating two different profiles from the SAME CC photo and be a little sloppy about when marking the corner patches, and see if you get different numbers applying those two profiles.  An idea where things might not be repeatable, is that there are slightly variations in the color of the color patches (you should be able to move the eyedropper across the color patch and see if the RGB numbers change) due to slight color noise and depending on where you put the "corner" markers on the CC image, you'll get slightly different results. 

             

            Does the exposure make any difference?  You can determine this by taking a photograph using the SAME lens in the SAME lighting (a few seconds apart), and just varying the exposure by 1/2 or 2/3 of a stop, and then computing a profile for each exposure and apply those two profiles to one of the exposures and see if the non-L coordinates of HSL or Lab eyedroppered. 

             

            If you check all these variations you'll have an idea of how much each affects the profile and then can judge if the magnitude of the differences you're seeing are related to variations with creating the profile, or actually related to differences in the lenses and thus a new profile for each lens might be warranted, assuming you can tell the difference, still.  I mean even if you can tell the difference between the profiles created with different lenses, are the differences from the lens significantly more than the differences due to exposure or lighting color or corner-patch placement?

             

            --

             

            I haven't tried computing a profile for each lens; however, I have created a dual-illuminant profile (2700K and 6500K) and then computed new color-matrix slider values (the ones under where you set the profile) for various lighting conditions using Tindemans' script and despite the slider values being not close to zero, I can hardly tell any difference on the few images I've looked at.  Once exception to not having the color-matrix sliders make much difference is when using the dual-illuminant profile with fluorescent lighting, which has a significant Tint value compared to either of the standard illuminants, but in the case of fluorescent lighting, I'd rather compute a whole new profile, than use a slider-corrected dual-illuminant profile.

             

            --

             

            Besides eyedroppering Lab or HSL coordinates in Photoshop, another way to check for color variations is to create a color-error plot in the Color Check module of Imatest and see how far the squares and circles are off from each other for each color-patch.  An example of such a color-error plot is linked below, where it shows how far off the colors of a color-checker are in incandescent lighting after computing a color-profile in incandescent lighting.  You'd expect them to be completely correct, but they aren't, and is a lesson in color profiles only being to go part way in making the colors look as if they were photographed in sunlight:

            http://www.pbase.com/ssprengel/image/101322979

             

            If you click on the above image, you will return to the thumbnails for color-error the gallery, and in the gallery description you can see links to both Imatest and Tindemans' script if you care to pursue things more in depth.  Imatest is not free but does have a free 30-day trial, which should be enough time to get some useful information out of it.

            1 person found this helpful
            • 3. Re: CCP colour profiles and different lenses
              Mollysnoot Level 3

              Many thanks to you both for your expertise, I very much appreciate your input.

               

              I'll follow up on some of the information supplied, and also do a little more testing, but on the face of it it seems that I won't need to create multiple profiles for each and every lens; this is actually something of a relief, as I was concerned I'd end up with a very cumbersome list of profiles in ACR/develop settings in Bridge!

               

              • 4. Re: CCP colour profiles and different lenses
                Noel Carboni Level 7

                I am but an egg by comparison to the posters above, but for what it's worth...

                 

                In the past I have gotten good results from making a single profile for a single camera model, then fine tuning using saved default slider settings to get the color I want as a default starting point.  At one point not long ago Adobe released updated profiles, and I re-evaluated...  At the moment I find the Adobe-supplied Camera Standard profile an excellent starting point for this strategy.  This seems almost overly simplistic by comparison to what you were contemplating, but it does actually work.

                 

                -Noel

                • 5. Re: CCP colour profiles and different lenses
                  Hudechrome Level 2

                  Be careful not to get some of that egg on your face, Noel!

                  • 6. Re: CCP colour profiles and different lenses
                    Marco N. Level 1

                    Last year I made a detailed study on these topics, you can read it here.

                    To summarize the findings and contextualize them, lenses with a
                    so different spectral transmittance that causes a temperature difference of 450 Kelvin carry out a relative color difference not detectable by the human vision system (on digital system not films), so just one color profile for these lenses is sufficient.

                     

                    Are your lenses more different?

                    • 7. Re: CCP colour profiles and different lenses
                      Hudechrome Level 2

                      Interesting test, Marco. Thanks for the link.

                       

                      I haven't read the entire work yet, but one thing caught my eye. In step 4, you show the exposure values with a correction of -1. Why -1 and not 0?

                      • 8. Re: CCP colour profiles and different lenses
                        Noel Carboni Level 7

                        Lawrence, its possible he's found that highlights are clipped at 0 EV and negative exposure compensation retains them as very bright but not maxed-out.  My own Camera Raw Default settings, used with the Adobe-supplied profile, contain a -0.25 EV exposure compensation for precisely this reason.

                         

                        -Noel

                        • 9. Re: CCP colour profiles and different lenses
                          Hudechrome Level 2

                          Playing off the clipping point against increased shadow noise is a dance I continually play. Complicate that by the fact that, if choosing A or S for Priority mode, a slight shift in camera aiming can cause a significant change in the EV.

                           

                          My compensation depends on the lens. For my general lens, it's +0.33EV. My 70 to 300 has an error that shifts the exposure further into the + side. I may have to send it back to Nikon.

                          • 10. Re: CCP colour profiles and different lenses
                            Marco N. Level 1

                            Hudechrome wrote:

                             

                            Interesting test, Marco. Thanks for the link.

                            I haven't read the entire work yet, but one thing caught my eye. In step 4, you show the exposure values with a correction of -1. Why -1 and not 0?

                            Thank you to you. That screenshot has only the purpose to tell that even the exposure has an influence on final colors (lightness), -1 is not an exposure correction in ACR but an exposure compensation in camera (the target background is black): nothing is clipped in the images (whites are near RGB 245 ).

                            • 11. Re: CCP colour profiles and different lenses
                              Hudechrome Level 2

                              Yes, I read it as an exposure correction in camera, as mine show up, and not in ACR What you showed was the camera data which doesn't change because of ACR adjustments..

                               

                              I know full well about the influence on color, exposure values have. Further, even differences in dynamic range of the subject camera has an influence, which I use when shooting, as the D90 can shoot at least to ISO 800 with smooth results, if correctly noise corrected. ISO 100 is reserved for those situations requiring the widest dynamic range possible.

                               

                              So I stll ask why that correction was used in the process. How would the results differ if you applied no adjustment? By differ, I don't mean in the obvious sense visually, but in the measurement sense as to what you are trying to achieve.

                               

                              Thank you

                              • 12. Re: CCP colour profiles and different lenses
                                Marco N. Level 1

                                Hudechrome wrote:


                                So I stll ask why that correction was used in the process. How would the results differ if you applied no adjustment? By differ, I don't mean in the obvious sense visually, but in the measurement sense as to what you are trying to achieve.

                                 

                                I have problems with English (but Google translator help me) so I hope to understand right the topics.

                                 

                                The exposure compensation is due to how metering system works: the color background behind the target is black that induces the metering system to push on exposure time (the time to bring the measured zone around lightness L*=50). I used the aperture mode via remote control and setting in camera a exposure compensation of -1 stop I obtained in ACR the white patch around RGB 245.

                                • 13. Re: CCP colour profiles and different lenses
                                  Hudechrome Level 2

                                  Ok, Makes perfect sense. I do that outdoors as well.

                                   

                                  Thanks, Marco.