18 Replies Latest reply on Mar 28, 2011 5:04 AM by Marco N.

    RGB curve and saturation

    Marco N. Level 1

      Why ACR doesn't use a luminosity curve to avoid the saturation boost of the RGB curve?

       

      Thank you

      Marco

        • 1. Re: RGB curve and saturation
          Jeff Schewe Level 5

          Because in testing, Thomas Knoll decided that a slight saturation boost (without Photoshop's hue change) was more natural. He tested both and made the decision...so you'll have to take it up with him.

           

          The odds are, you would STILL need to tweak saturation after adjusting curves even if the curves were luminance only. The way it is now, a slight reduction of overall saturation will eliminate the saturation boost caused by the curves.

          • 2. Re: RGB curve and saturation
            Marco N. Level 1

            And this is the reason because the custom profiles are often too much saturated, right?

             

            What method do you suggest to tweak the saturation after contrast boost? Saturation slider, Vibrance, HSL slider or directly by profile tweak (DNG PE)? What are the differences between these methods?

             

            Thank you for your answers

            Marco

            • 3. Re: RGB curve and saturation
              Jeff Schewe Level 5

              Marco N. wrote:

               

              And this is the reason because the custom profiles are often too much saturated, right?

               

              No...the DNG Profiles control the hue rendering more than saturation...how warm or cool the reds are as an example. Yes, the DNG Profile has an impact on saturation but the biggest differences are really the hue rendering.

               

              Marco N. wrote:

               

              What method do you suggest to tweak the saturation after contrast boost? Saturation slider, Vibrance, HSL slider or directly by profile tweak (DNG PE)? What are the differences between these methods?

               

              Depends...a curves tweak generally has an across the board impact on saturation. Increase the contrast and the saturation increases and decreases when you do a contrast reduction. So, the "best way" really depends on your image.

               

              If the saturation is bumped up too high for a given hue, then I would suggest HSL for the specific hue/sat combo. For an overall saturation reduction the main Saturation slider is prolly better than Vibrance.

               

              Tweaking profiles would only be suggested if your particular camera rendering is different or if the light source impacts your sensor in an odd manner. Tweaking profiles isn't really a color correction technique, it's more a sensor response tweak.

              • 4. Re: RGB curve and saturation
                Marco N. Level 1

                >No...the DNG Profiles control the hue rendering more than  saturation...how warm or cool the reds are as an example. Yes, the DNG  Profile has an impact on saturation but the biggest differences are  really the hue rendering.

                 

                Both calibration scripts and  DNG PE calculate the right saturation  in a situation not close to the setup  of real world images (low contrast to match the lightness values of gray  patches of CC / high contrast of real word images to get close what the  eyes see). The raising of the contrast in real world images sould be  the reason of the oversaturation of the custom profiles in their use

                 

                What do you think?

                 

                Thank you

                Marco

                • 5. Re: RGB curve and saturation
                  Jeff Schewe Level 5

                  Sorry...I don't really understand your question.

                   

                  What is your point?

                  • 6. Re: RGB curve and saturation
                    Marco N. Level 1

                    The starting point is that I see the DNG profiles customized by both DNG PE and calibration script at ACR default render the real world images oversaturated and for curiosity I ask me what is the cause.

                     

                    Thank you

                    Marco

                    • 7. Re: RGB curve and saturation
                      Jeff Schewe Level 5

                      Well, I don't agree that Camera Raw renders image as over saturated...I tend to use Adobe Standard with my images from Canon cameras but use a custom DNG for my Phase One captures. In both cases, I see the saturation at normal default for be rather accurate. Far more accurate than the vender matching DNG profiles.

                       

                      So, without more info regarding your camera model and exactly how you made the DNG profiles (single or dual-illuminate) there's not much I can comment on other than I don't agree with your assessment...

                      • 8. Re: RGB curve and saturation
                        Vit Novak Level 3

                        Marco,.

                         

                        All profiles have a tone curve (if not, default tone curve is used), which is compressing relatively large dynamic range of the sensor into much smaller dynamic range of output media / output color space. Because of this curve, shadows and highlights are compressed.

                         

                        However, this tone curve is matched to some average scene. If your scene has more or less dynamic range than that, you will find that your photo can look whether oversaturated or undersaturated. Also, on the same photo, you will find that some colors look oversaturated, but the same colors, just brighter or darker, look undersaturated, depending how compressed they are with a tone curve.

                         

                        So it's hard to make any statement that some profile is accurate or that it isn't, because what you see on the monitor or paper is only a "painting" of reality and will never look the same as reality as long your scene has more dynamic range than your output media

                         

                        And another thing, with calibrating chart consisting of 18 color patches you can only expect that only exactly those colors will be accurate and noting more

                        • 9. Re: RGB curve and saturation
                          Noel Carboni Level 7
                          function(){return A.apply(null,[this].concat($A(arguments)))}

                          Vit Novak wrote:


                          what you see on the monitor or paper is only a "painting" of reality


                          That is SO VERY true, and so few people really grasp it as a concept.

                           

                          -Noel

                          • 10. Re: RGB curve and saturation
                            Marco N. Level 1


                            >Thomas Knoll decided that a slight saturation boost (without Photoshop's hue change) was more natural.or in

                             

                            Today I made some proofs but I see the same result applying a curve in Photoshop or in Camera Raw: use the google translator if you want read the article: http://www.gialandra.it/blog/files/27771d380fddb6f206a7bbcff06c5de9-6.html

                             

                            >So, without more info regarding your camera model and exactly how you made the DNG profiles (single or dual-illuminate) there's not much I can comment on other than I don't agree with your assessment...

                             

                            I made DNG profile not only for one camera and in all ways, by calibration scripts, by DNG PE and using a combination of both methods, at one and dual illuminant. In the effort I use a i1 Pro spectro and SoLux bulbs with an adjustable power supply voltage (CRI 97 with D50). With the generated profiles the average deltaE 2000 are in the order of the unit from spectro values.

                             

                            I use the ColorChecker Passport too but the results are very close.

                             

                            I resolve this problem via DNG PE and the results are good, but I want to know if someone else has feel this problem and understand the cause of it.

                             

                            >So it's hard to make any statement that some profile is accurate or that it isn't, because what you see on the monitor or paper is only a "painting" of reality and will never look the same as reality as long your scene has more dynamic range than your output media

                             

                            So, if I open an image and at default it has the tone mapping correct, the dynamic range should not be a problem, and the image should be correct even in saturation … is this right?

                             

                            I agree that my monitor provides its vision of what there is write in the image file.

                             

                            >And another thing, with calibrating chart consisting of 18 color patches you can only expect that only exactly those colors will be accurate and noting more

                             

                            This is true if you use only DNG PE. If you use a good calibration script before using the DNG PE even the colors far from sampling points are enough good and you have other benefits.

                             

                            Thank you everybody

                            Marco

                            • 11. Re: RGB curve and saturation
                              Jeff Schewe Level 5

                              What calibration script are you talking about?

                              • 12. Re: RGB curve and saturation
                                Jeff Schewe Level 5

                                Marco N. wrote:

                                 


                                >Thomas Knoll decided that a slight saturation boost (without Photoshop's hue change) was more natural.or in

                                 

                                Today I made some proofs but I see the same result applying a curve in Photoshop or in Camera Raw: use the google translator if you want read the article: http://www.gialandra.it/blog/files/27771d380fddb6f206a7bbcff06c5de9-6. html

                                 

                                 

                                I read the translated article (so I don't know exactly how good the translation was) but you seemed to have been using a relatively mild curves correction not a really strong one. It's on strong curves corrections that Camera Raw's "hue protected" will show as being better (different) than Photoshop's curves correction which will tend to twist hues as well as changing saturation.

                                 

                                I'm pretty sure (mainly since Thomas says so) that Camera Raw's curves work differently (better) than Photoshop's curves when it comes to altering the hue of a color when applied using a normal blend mode in Photoshop. Both curves functions have an impact in saturation.

                                 

                                So, while it's true that you can't do a luminosity based curves correction in Camera Raw (as some "experts" suggest is required), a curves correction and a slight saturation adjustment can accomplish the same effect on an image.

                                 

                                Again, this is based on what Thomas has said and as the coauthor of Photoshop and the founding engineer of Camera Raw I would tend to defer to Thomas in these matters. Camera Raw ain't Photoshop. Camera Raw was specifically designed (13 years AFTER Photoshop was released) to do optimal raw capture editing. There are things that can be done in Camera Raw that can't be done as well in Photoshop. But the upside is that Camera Raw works as an import plug-in FOR Photoshop, so anything you can't do inside of Camera Raw can easily be done afterwords in Photoshop. But I do suggest optimizing the image as best as possible inside of Camera Raw for optimal image quality.

                                • 13. Re: RGB curve and saturation
                                  Tim Lookingbill Level 1

                                  Marco, I get the same saturation boost using traditional S-curves as you describe.

                                   

                                  What I've found to fix this in ACR is to construct a contrast curve with flat sections in the tonal areas that show the saturation. It will depend on how extreme the tonal variances are within each individual image to determine where these flat sections are placed.

                                   

                                  Another alternative method is, after setting white/black clipping thresholds, set Contrast slider to zero and tone curve to Linear along with a profile that gives the most desired hues. After your eyes have adjusted to the flatness and less saturation, first adjust the black node in the corner of the Linear curve and slide it to the right to add contrast from the bottom end without blowing out endpoints. Go back and tweak Contrast and/or Black sliders.

                                   

                                  When I add contrast this way I notice there is less saturation. This may introduce tonally flat areas where then I add adjustment nodes to the linear curve and/or add Clarity. Fill slider usually adds saturation. I usually stay away from Recovery as much as possible.

                                   

                                  This worked great with high contrast shots of close up detail of snow scenes lit by morning sunlight where the shadows were quite dark with very saturated blues. Instead of solely relying on the contrast tweaks to reduce the saturated blue shadows, I just reduced saturation in the blue channel in HSL panel.

                                   

                                  I would suggest you approach Raw processing like molding a lump of clay using all the tools available.

                                   

                                  All things being the same in studio work it's possible to turn the camera into a copier for image creation by the numbers, but that only works in a controlled environment. Once you shoot under different lights and angles (outdoors) you'll have to forget the numbers and start all over.

                                   

                                  As you know the defaults are only a starting point and now that you know how saturation increases with contrast you'll know what tool and method to adjust to control this. It's doable in ACR, believe me. I correct for this saturation/contrast issue all the time to where all I do in Photoshop is sharpening and resizing for web and/or print.

                                  • 14. Re: RGB curve and saturation
                                    Vit Novak Level 3

                                    As mentioned by Jeff, curves in ACR and Photoshop don't work exactly the same way

                                     

                                    In Photoshop, curve is applied in your working space (which is nonlinear output color space) to individual R, G and B component. So if curve is not linear, it changes hue, saturation and, of course, lightness. In ACR, tone curve is applied to brightest and darkest of these tree colors. Third color is calculated so that hue remains constant. So, tone curve in ACR changes saturation and lightness. But It works in linear color space - it's another difference. So applying the tone curve in ACR and what you think is the same tone curve in Photoshop (althoug it's not the same) will produce slightly different result

                                     

                                    However, it should be noted that curve shown in ACR is NOT actual tone curve that is applied to raw image. You can see actual tone curve in DNG PE (there is a check box for this I think) and it has considerably raised middle part, so it is brightening the image and compressing the highlights, affecting saturation. Shape of actual tone curve is controlled by sliders brightness, contrast and with tone curve shown in ACR. If you set it to default (brightness 50, contrast 25, middle contrast tone curve), you will get default Adobe tone curve, shown in DNG PE

                                     

                                    With camera color profiles, whole thing about actual tone curve is even more complicated

                                     

                                    So without knowledge how this actually works, above blog, mentioned by Marco, doesn't make much sense to me - it only shown that curves don't work the same way in Photoshop and ACR, as results are different. Also, I don't se the reason to use "linear tone curve" in ACR and applying tone curve in PS, because effective tone curve is not linear. It becomes linear if you also set brightness and contrast to 0 (this is true only for profiles with default Adobe tone curve)

                                    • 15. Re: RGB curve and saturation
                                      Marco N. Level 1

                                      >What calibration script are you talking about?

                                       

                                      Tindemans script, I have studies all three script but the precision of this script is surprisingly.

                                       

                                      >you seemed to have been using a relatively mild curves correction not a really strong one

                                       

                                      I have tried again now the test (linear+PS vs strong contrast, it need less than 5 minuts now that I have the calculated PS curve) and the results is the same as before. But I suspect why Tomas say so, I want make another test before and after we can see if I have guess right.

                                       

                                      >So, while it's true that you can't do a luminosity based curves correction in Camera Raw (as some "experts" suggest is required), a curves correction and a slight saturation adjustment can accomplish the same effect on an image.

                                       

                                      I'm agree, this is not a big problem (for me there are other bigger problem in ACR). But why don't accomplish this effect directly in DNG PE once for all? (I use a little de-saturation of the profile)

                                       

                                      I have forgot to tell that I found the same problem with ICC profiles.

                                       

                                      >But I do suggest optimizing the image as best as possible inside of Camera Raw for optimal image quality.

                                       

                                      Here in Italy there is a promising photographer that sustains strongly the "all in PS school" on your very frequented forum. He reset to zero all sliders and even the curve to linear... I'm agree with you (I arrive from Bruce's school), and on your forum I sustain that he lose quality and many TIME.

                                       

                                      >What I've found to fix this in ACR is…

                                       

                                      Tim, as report above a desaturated profile are time and frustration saving for me. Thank you for your report

                                       

                                      >Once you shoot under different lights and angles (outdoors) you'll have to forget the numbers and start all over.

                                       

                                      In the past (if you want I give you the link) I see that a well calibrated dual-illuminant profile works very well for a long range of WB (natural and tungsten lightning) from 2500K to over 10000K. Obviously if the SPD of the light goes away from these conditions a new profile is needed.

                                       

                                      >So if curve is not linear, it changes hue, saturation and, of course, lightness.

                                       

                                      Vit, I suspect that the what keeps the hue still is not the gamma working space of ACR, but another characteristic of the working spaces. Bruce Fraser once explained this stuff and I want verify his words in the test cited above.

                                       

                                      The blog that I have mentioned is my blog.

                                       

                                      Thank you for your interventions

                                      Marco

                                      • 16. Re: RGB curve and saturation
                                        Vit Novak Level 3

                                        Marco

                                         

                                        There is no such thing as "gamma working space of ACR". Color transformations in ACR happen mostly in linear photo pro color space (although Eric is also considering alternative approach because of some drawbacks of this method), so tone curve is applied in that space. Conversion to output color space (if it's not Photo Pro) and gamma encoding happens at the end of the workflow

                                         

                                        Thing that keeps hue constant when tone curve is applied in this workflow is a part of ACR code that is programmed that way and documented in dng sdk. I studied part of that code when making my program for generating camera dng profiles for cameras that are not supported by ACR. It's programmed that way by desing and there is nothing particulary wrong (or good) about it. However, dng sdk documents only a part of ACR code. For instance, it doesn't tell how exactly tone curve that is shown in ACR is merged with the tone curve in the profile and the way how brigtness, contrast and blacks sliders modify this curve.

                                        • 17. Re: RGB curve and saturation
                                          Tim Lookingbill Level 1

                                          Still not sure what is being accomplished in this topic.

                                           

                                          Are you trying to reverse engineer ACR's tools as a way to fix a saturation/contrast issue?

                                           

                                          I've been tinkering around with ACR for several years and I don't find there's a problem.

                                           

                                          Most of the problems I find shooting Raw has mostly to do with quality and quantity of light used in front of the camera. There are some really strange spectral radiance influenced characteristics involving mostly optical effects that I can bring out or reduce using ACR, but discovered it's the light that's causing it and not the software.

                                           

                                          I've been studying all the variety of lights available both artificial and natural from CFL's to regular living room incandescent for photography and what it does to pixels that not even a DNG profile can fix. It's been an eye opener for me from a research standpoint.

                                           

                                          Thought maybe this thread would offer other insights I hadn't come across, but as of yet I haven't read anything that hasn't already been covered.

                                           

                                          So really, what is being resolved here?

                                          • 18. Re: RGB curve and saturation
                                            Marco N. Level 1

                                            I have finished this test (it should be enogh to see only the images to understand it) even if it is started from a point of view and finished in other. I have tested the hue behavior of the standard working spaces plus ACR applying a strong tonal curve (same strengh as from linear to strong curve in ACR). It appears that ProPhoto in PS is the best color space in term of keeping hue costant, followed by ACR.