10 Replies Latest reply on Jan 4, 2013 5:24 AM by Romain_Th

    Highlights / High saturation posterization still not resolved with PV2012

    Romain_Th Level 1

      Hi,

       

      IMO ACR's rendition around highlights has always been its week point. This has been greatly improved with PV2012 and with v4 profiles (including compensation for the BaselinExposure tag), but their are still situation were nothing can be done:

       

      This is from a D800 using the Adobe Standard profile (please note that using Camera Standard, the rendition is better but not perfect either). There is a harsh transition from white to yellow, then from yellow to orange, then from orange to the wall color:

      DSC_3391_LR.jpg

      Then, the same rendered with Capture One 7, much smoother:

      DSC_3391_CO.jpg

      This example is pretty obvious (rendered using the Camera Standard profile, the best profile):

      DSC_1401_LR.jpg

      And the same from Capture One 7:

      DSC_1401_CO.jpg

      I tried to modify the different profiles using the Adobe tool, but it is too imprecise. I also tried to modify the different profile curves to smooth them but without success (using DCP tool).

      It seems the problem comes from the fact that the various LUT tables in the profiles are applied after white balance, excluding clipped data.

       

      Any help with this? Or any update to come? I really prefer LR's workflow from C1!

       

      Romain

        • 1. Re: Highlights / High saturation posterization still not resolved with PV2012
          Yammer Level 4

          I can't see the posterisation in the first example, although the region around the lamp seems blown out, and there is less recovery in the CaptureOne rendering.

          The second example shows clear posterisation, with a crimson halo around the sun. What concerns me with this image is the large negative Tint value and the sRGB colour space, which is going to struggle with the reds in a sunset.

           

          Any chance of the raw files, from webspace or YouSendIt, Dropbox, etc.? If we can see the images on our own systems, we might get a better idea of the problem.

           

          PS. What is a "Generic RGB Profile" (which is what the ICC profile is named in the EXIF metadata)?

          • 2. Re: Highlights / High saturation posterization still not resolved with PV2012
            Romain_Th Level 1

            Transitions in the first example are bad, you can see the jumps between the different hues. I agree that there are less recovery in the C1 example, so I added the same with some recovery. You can see that it is perfectly smooth.

             

            The most annoying thing is probably the orange halo. You can see it in the first example but it is not very obvious. It is however, on the second ond.

             

            It is not a colour space issue since I can see the exact same problem inside Lightroom (prophoto).

             

            You may download the raw files here: https://www.dropbox.com/sh/j01hoomjfdxzf34/ckoM36ONxr

             

            PS: "Generic RGB Profile" is the equivalent for sRGB in C1.

             

            C1 with some recovery:

            DSC_3391.jpg

            • 3. Re: Highlights / High saturation posterization still not resolved with PV2012
              Yammer Level 4

              I've just spent the last half-hour re-familiarising myself with Rawnalyze and playing with your NEF files.

               

              The bedroom scene's highlights are blown in all three Raw channels to varying degrees. Using highlight recovery exaggerates the boundaries between clipped channels in ACR--not normal posterisation as such. No doubt this could be improved, but you are asking a lot of the developers. You can't make a silk purse from a sow's ear.

               

              Rawnalyze, showing clipped raw channels

              DSC_3391-R.jpg

               

              ACR (0 highlight recovery)

              DSC_3391-YP.jpg

               

              ACR with highlight recovery (eurgh)

              DSC_3391-YP-HR.jpg

               

               

              The sunset scene was well-handled, with only the sun's disk being blown out in the red and green channels. However, as I mentioned earlier, the deep red/orange hues are outside the gamut of sRGB, and so the image needs to be manipulated until the colours fit.

               

              Rawnalyse, showing only the disk clipping

              DSC_1401-R.jpg

               

              ACR with large amounts of highlight recovery to protect the deep hues around the sun before conversion to sRGB.

              DSC_1401-YP.jpg

              You can see that the disc and its reflection are posterised, but this is almost acceptable given the subject matter.

              • 4. Re: Highlights / High saturation posterization still not resolved with PV2012
                Romain_Th Level 1

                Thanks for your analyse!

                 

                Regarding the first example, I agree the initial data is pretty bad. I am probably asking a lot to the dev, but as I mentionned, this is really IMO the weak point of ACR because if you look at Aperture, CaptureOne or Capture NX, they will all have better transitions in the highlights.

                The problem appears quite often since a lot of pictures contain specular highlights (that are burnt). It is usually OK when it is pure white but saturated car lights are wrongly rendered.

                 

                Regarding the second example, I am not sure I understand, since you can see the problem in LR in pro photo color space, or you can export in a large color space and see it is still here. I agree that using massive amount of recovery may mitigate the problem, but still, other software are rendering properly without anything. Plus, when using recovery, you are darkening the highlights which make the image less interesting. C1 achieve way higher saturation without halo (in sRGB).

                • 5. Re: Highlights / High saturation posterization still not resolved with PV2012
                  Yammer Level 4

                  What you are asking for (in the bedroom photo) is for better interpretation of missing data due to channel clipping. Posterisation is a result of highlight recovery on guessed data. The comparison with the Capture One photo is unfair, as no recovery is attempted and the image is rendered brighter.

                   

                  I don't understand why you see banding in Lightroom. I don't use Lightroom. In Camera Raw (which uses the same 'engine'), I see banding when I pick a smaller colour space and make no attempt and reducing the image saturation/brightness to fit. Natural subjects like flowers and sunsets just don't fit into sRGB, and you need to make special exceptions to these subjects IF you have to fit them into sRGB. Using AdobeRGB or ProPhoto causes less problems, as the histograms are 'wider' and can fit the colour extremes more easily.

                   

                  Fitting wide gamuts into small profiles is ugly, and quite possibly other converters do a better job.

                  • 6. Re: Highlights / High saturation posterization still not resolved with PV2012
                    ssprengel Adobe Community Professional & MVP

                    The sunset in ACR and LR have identical posterization boundaries on my system.  Yammer do you have a wide-gamut monitor, while Romain do you have a standard gamut monitor?

                     

                    The area of posterization boundary in LR and ACR on my system is where the photo’s colors go out-of-gamut with respect to my monitor’s color profile.

                     

                    I can check this using LR4, by enabling Soft Proofing with the checkbox just above the thumbnail strip at the bottom, then changing the soft-proofing Profile just below the histogram area to ProPhotoRGB (a wide profile) and enabling the Monitor out-of-gamut warning (click the monitor symbol on the lefthand side of the histogram), or changing the soft-proofing Profile to my monitor’s profile and enabling the paper out-of-gamut warning, shown below.

                     

                    Without a wide-gamut monitor, I can’t test my theory but perhaps C1 is converting to the least-common-denominator of the monitor’s profile or the output profile whichever is smaller whereas Adobe isn’t converting to the monitor’s profile at all and has a large working color-space so there is an ugly transition when the colors go out of gamut on the monitor.

                     

                    I say this because In Romain’s C1 version, the area above the sun, that is within the posterization boundary in Adobe, is significantly more yellow indicating a hue-shift caused by the red-channel clipping and the green-channel becoming proportionately more equal (red+green=yellow).  This yellow has to be coming from somewhere that is not in the original raw data because the red-channel isn’t clipped except for the sun’s disk and its reflection, and perhaps caused by C1 color-engine fixing the out-of-monitor-gamut situation on the display.

                     

                    It would be interesting to test the sunset JPG exported out of C1 using the ProPhotoRGB colorspace (if that’s possible) then import it into LR or PS and check for gamut warnings with respect to sRGB.  I don’t have C1 so can’t test this theory, either.

                     

                    The bottom line is that Adobe's display of a wide-gamut image looks bad on a narrow-gamut monitor because LR and ACR don’t try to fix things for the monitor’s colorspace—that is left up to the user to do, by reducing the brightness or manually shifting the hue.

                     

                    Here is a screenshot of my LR4's soft-proofing showing the out-of-gamut warning for my monitor profile:

                    2013-01-03_180658.jpg

                    • 7. Re: Highlights / High saturation posterization still not resolved with PV2012
                      Yammer Level 4

                      Yes, of course, I'd forgotten all about my monitor. I have an almost-AdobeRGB wide gamut monitor, and that explains why I'm not seeing any posterisation on wide-gamut images. I normally process photos in the AdobeRGB colour space, so I normally see what I get (if you know what I mean).

                       

                      I get a similar problem to that described by the OP when browsing images in Bridge (my main workflow starting point). Bridge uses ACR to generate previews in sRGB (Relative Colormetric). So, when I preview an unprocessed wide-gamut Raw image in Bridge, I sometimes see clipped colour effects (although I must say that, in this case, they are still not as bad as that seen in the OP's sunset JPEG).

                       

                      The two example photos demonstrate two different problems. The first being highlight recovery with clipped channels; the second being display of wide-gamut images, as well as clipped channels (in the sun's disc and reflection).

                       

                      It appears that CaptureOne converts to the monitor's profile, maybe using Perceptual intent (hence the orangey appearance). This will give a less distorted image preview, but this ultimately has no effect on the finished image—indeed it gives the user a false sense of security that all is okay if they forget to soft-proof.

                      • 8. Re: Highlights / High saturation posterization still not resolved with PV2012
                        Vit Novak Level 3

                        Recently I've been doing some experiments about this problem in ACR, connected with highlights. My conclusion is - this is the weak point of ACR. However, it's the weak point only if Adobe standard profile is used. With custom made profile, it's possible to avoid this. But, it's not that simple to make one

                         

                        Main reason for highlight issues is a very simple method of gamut mapping in ACR workflow. Namely, after matrix conversion from raw to output color space, some colors are out of gamut. ACR maps them into gamut by simply clamping the saturation and value (in HSV) into range 0..1, and this method sometimes performs very bad. Another problem is applying a tone curve in Photo Pro. Its color gamut is much bigger than Adobe RGB or sRGB and applying a tone curve there throws some colors way out of gamut of Adobe RGB / sRGB, resulting in big errors when they are finally mapped into Adobe RGB / sRGB by simply clamping RGB values. This is mostly visible in blue range, where some hues look way oversaturated/blown/wrong (like dark sky color). We already discussed that on sample from Olympus camera

                         

                        With custom profiles, gamut mapping and transition into highlights can be controlled preciselly and camera profiles generally perform much better in this area, because camera manufactures used better gamut mapping method. During last decade, there was some research about doing it properly, even some books published, but even from latest DNG SDK, nothing seems to be changed in ACR workflow lately (actually, there are some changes about dealing with highlights, which are not documented in SDK, but it's not enough)

                        • 9. Re: Highlights / High saturation posterization still not resolved with PV2012
                          Romain_Th Level 1

                          Hi,

                          What you are asking for (in the bedroom photo) is for better interpretation of missing data due to channel clipping. Posterisation is a result of highlight recovery on guessed data. The comparison with the Capture One photo is unfair, as no recovery is attempted and the image is rendered brighter.

                           

                          I don't think it is unfair when you look at the second example of the bedroom. But yes I am asking for better interpretation around clipped data. Other raw converters seems to do it pretty well.

                           

                          The sunset in ACR and LR have identical posterization boundaries on my system.  Yammer do you have a wide-gamut monitor, while Romain do you have a standard gamut monitor?

                           

                          Yes I have a standard gamut monitor. And yes the issue is due to the monitor clipping at high saturation.

                           

                          It would be interesting to test the sunset JPG exported out of C1 using the ProPhotoRGB colorspace (if that’s possible) then import it into LR or PS and check for gamut warnings with respect to sRGB.  I don’t have C1 so can’t test this theory, either.

                           

                          The same image rendered in ProPhoto is exacly the same as the one in sRGB, without any clipping. I guess the converter is volontary keeping the colors into a "secured" area. I can boost the saturation a lot before having to export in a colorspace bigger than srgb. Is this due to the camera profile in C1? I have no idea, and developping without camera profile is still nowhere in the clipping area of srgb.

                           

                          The two example photos demonstrate two different problems. The first being highlight recovery with clipped channels; the second being display of wide-gamut images, as well as clipped channels (in the sun's disc and reflection).

                           

                          Yes and no: IMO, ACR try to keep saturation high in highlights (probably because it is working a lot in HSL) whereas in C1 the saturation decreases in highlights and hues are shifted, allowing smoother rendition without clipping in most colorspaces. (I agree the sunset's halo is due to the monitor profile beeing to small though)

                           

                          It appears that CaptureOne converts to the monitor's profile, maybe using Perceptual intent (hence the orangey appearance). This will give a less distorted image preview, but this ultimately has no effect on the finished image—indeed it gives the user a false sense of security that all is okay if they forget to soft-proof.

                           

                          I was using soft proofing in C1. And the JPEG posted is a rendered one in sRGB, so their is no false sense of security: I saw in the viewer what I got in the final JPEG which is not clipped in sRGB. But even when changing the proofing profile in C1, the colors don't change because in this image, the colors fit sRGB (in C1 but not in ACR).

                           

                          Recently I've been doing some experiments about this problem in ACR, connected with highlights. My conclusion is - this is the weak point of ACR. However, it's the weak point only if Adobe standard profile is used. With custom made profile, it's possible to avoid this. But, it's not that simple to make one

                           

                          Main reason for highlight issues is a very simple method of gamut mapping in ACR workflow. Namely, after matrix conversion from raw to output color space, some colors are out of gamut. ACR maps them into gamut by simply clamping the saturation and value (in HSV) into range 0..1, and this method sometimes performs very bad. Another problem is applying a tone curve in Photo Pro. Its color gamut is much bigger than Adobe RGB or sRGB and applying a tone curve there throws some colors way out of gamut of Adobe RGB / sRGB, resulting in big errors when they are finally mapped into Adobe RGB / sRGB by simply clamping RGB values. This is mostly visible in blue range, where some hues look way oversaturated/blown/wrong (like dark sky color). We already discussed that on sample from Olympus camera

                           

                          With custom profiles, gamut mapping and transition into highlights can be controlled preciselly and camera profiles generally perform much better in this area, because camera manufactures used better gamut mapping method. During last decade, there was some research about doing it properly, even some books published, but even from latest DNG SDK, nothing seems to be changed in ACR workflow lately (actually, there are some changes about dealing with highlights, which are not documented in SDK, but it's not enough)

                           

                          These are also my findings, though even the "Camera" profiles developped to match the manufacturer's rendition are not perfect in highlights and there is no way to modify the matrix easily.

                           

                          IMO, ACR is lacking an option to compress saturation levels smoothly without hard clipping (like a perceptual intend ICC convertion). This would fix color problem around burnt areas / and or very saturated levels. And this is very common arount specular lights or in saturated lighting conditions (sunset / concerts...).

                          • 10. Re: Highlights / High saturation posterization still not resolved with PV2012
                            Romain_Th Level 1

                            For the record, below is an attempt to get the same clipping area in sRGB using C1. This required a 30% boost in saturation in the yellow/orange/red colors. Notice how the image is way more saturated the ACR's one and the transition is still smoother. But here your can see the halo at the non-clipped to clipped transition.

                             

                            DSC_1401.jpg