20 Replies Latest reply on Jan 27, 2012 4:30 PM by olgol

    The meaning of "contrast" in ACR 4.1

    Panoholic Level 2
      What is the meaning of negative contrast?

      What is the setting for "do not apply any contrast adjustment"? Zero or -50?
        • 1. Re: The meaning of "contrast" in ACR 4.1
          Thomas Knoll Level 2
          I don't understand the meaning of the question: "do not apply any contrast adjustment".

          From what baseline are you adjusting from? With raw files, there is no real fixed baseline.
          • 2. Re: The meaning of "contrast" in ACR 4.1
            Panoholic Level 2
            The "baseline" is the result of the must adjustments:

            - black level correction (not the "blacks" in ACR, but the substraction of the "dark current" values),

            - linearization if the original values are not linear,

            - applying the/some white balance

            What else?
            • 3. Re: The meaning of "contrast" in ACR 4.1
              Thomas Knoll Level 2
              Cameras capture scene referred data. Finished images are output referred. To get between the two, you must have some kind of non-linear tone mapping applied, or the resulting images will look flat and ugly. Is that what you want (scene referred data--which is ugly to look at directly)?

              Why are you asking these questions? If I knew the point of these questions, I would be better able to help you.
              • 4. Re: The meaning of "contrast" in ACR 4.1
                Level 1
                If you want to see a linear mapping of your image, set the Exposure & Blacks (4.x) to zero, set Brightness tp -150 & Contrast to -50 and set the point curve to linear...otherwise, the image will have some tone mapping...
                • 5. Re: The meaning of "contrast" in ACR 4.1
                  Panoholic Level 2
                  > you must have some kind of non-linear tone mapping applied, or the resulting images will look flat and ugly

                  The tone mapping is not standardized, but ACR too generates the mapping based on some algorythm, like CIE L* or Rec. 709 (gamma 2.2222). The original algorythms don't include a "redistribution" from the middle towards the two ends (high contrast) or from the ends towards the middle (low contrast), like following three histograms from the same image show, with contrast +25, +100 and -50:







                  What I am looking for is the setting, which leads to an "unadultered" mapping, as a starting point for adjustments.
                  • 6. Re: The meaning of "contrast" in ACR 4.1
                    Bill_Janes Level 2
                    > If you want to see a linear mapping of your image, set the Exposure & Blacks (4.x) to zero, set Brightness tp -150 & Contrast to -50 and set the point curve to linear...otherwise, the image will have some tone mapping...

                    The above settings will linearize the image, but also markedly darken the image. If you merely want to linearize the image, then setting Brightness to zero and Contrast to zero will produce a better result.

                    For illustration, here are photographs of Stouffer stepwedges with Contrast 0, Brightness 0 on the top, and Brightness -150 and Contrast -50 on the bottom:

                    stouf_composite

                    Here is the characteristic curve as plotted by Imatest of the top image. Note that the data are linear, but the gamma is not one.

                    ImatestPlot
                    • 7. Re: The meaning of "contrast" in ACR 4.1
                      Bill_Janes Level 2
                      G Sch wrote:

                      >The tone mapping is not standardized, but ACR too generates the mapping based on some algorythm, like CIE L* or Rec. 709 (gamma 2.2222). The original algorythms don't include a "redistribution" from the middle towards the two ends (high contrast) or from the ends towards the middle (low contrast), like following three histograms from the same image show, with contrast +25, +100 and -50:

                      I don't think that G Sch is using the correct tool for demonstrating the contrast curve of ACR.

                      Here are Imatest plots of the Contrast 50 and Contrast 0 renderings. As expected, the contrast 50 curve has a higher slope (gamma) in the mid portion and an S-curve has been applied to the two extremes.

                      Stouf_Composite

                      I don't know if G Sch knows who Thomas Knoll is when he is lecturing to him. FYI, Mr. Sch, Mr. Knoll is the creator of Photoshop and Camera Raw. In a previous thread with Mr. Sch, I found the discussion unproductive and broke off from the discussion. Your mileage may vary.
                      • 8. Re: The meaning of "contrast" in ACR 4.1
                        Level 1
                        "I found the discussion unproductive and broke off from the discussion."

                        Well, tut tut!

                        I'm not one to be very intestested in charts, graphs and Imatest but I am curious as to why G Sch wants to make such a rendering.
                        • 9. Re: The meaning of "contrast" in ACR 4.1
                          Panoholic Level 2
                          >I am curious as to why G Sch wants to make such a rendering

                          What rendering? All I asked for was some clarification of the nomenclature adopted in the ACR UI. I mentioned some algorythms of perceptually linear mapping as starting point for further adjustments.

                          Note, that the terms "increasing the contrast" and "decreasing the contrast" are misnomers. Adjusting the global contrast (in ACR the "contrast" setting among the basic adjustments) shifts the contrast, either towards the middle, or towards the extremes; the effect depends on the actual image. A "higher contrast" can result in reduced contrast with some images and vice versa. (I am not suggesting to change this usage of the terms, but one needs to be aware of this fact.)
                          • 10. Re: The meaning of "contrast" in ACR 4.1
                            Peter DL Level 1
                            G Sch wrote:
                            >> The tone mapping is not standardized, but ACR too generates the mapping based on some algorythm, like CIE L* or Rec. 709 (gamma 2.2222). The original algorythms don't include a "redistribution" from the middle towards the two ends (high contrast) or from the ends towards the middle (low contrast), ...<<

                            You are confusing gamma-encoding,
                            which is essentially irrelevant in a color-managed environment,
                            with the application of a visual effective, pleasing tone curve
                            (see post # 3 by Thomas Knoll).

                            Brightness and Contrast controls shape such pleasing tone curve, but they do not set the black and white endpoints (thats now Shadows and Exposure). You might wish to read about this in Bruce Frasers book. Brightness acts like a midtone slider and Contrast then applies an S-curve around the midpoint set by Brightness. Theres nothing basically wrong with this nomenclature. Its a kind of second generation definition for these terms. Note that Contrast isnt anymore a BP/WP-clipping slider as it formerly was in Photoshop. That said, the new definition of Brightness could probably be further improved by adding a proportional amount of new Contrast. Anyway.

                            If you want to skip this 'pleasing' tone curve, just set Brightness, Contrast and Shadows to zero. As already explained by Bill Janes (see post # 6) this yields a reasonable well linearization of ACR. However, final output to selectable standard working spaces is always gamma-encoded. So if you further want a numerically unaltered / linear mapping e.g. of a grayscale, convert the file back to a linear-gamma version of your working space. Note that this conversion does not change appearance.

                            In case of further questions - just ask (like you initially did when opening this thread) rather than raising more uninformed claims.

                            Peter

                            --
                            • 11. Re: The meaning of "contrast" in ACR 4.1
                              Panoholic Level 2
                              >You are confusing gamma-encoding, which is essentially irrelevant in a color-managed environment, with the application of a visual effective, pleasing tone curve

                              I am not confusing them, but I am not regarding the application of the gamma function irrelevant, because it does take place and contributes a big chunk to the resulting contrast.

                              I mentioned this, because the term "linear" have been used with different meanings in the preceding post, and it gave the impression, that the image were truely linear without applying the "pleasing tone curve".

                              >If you want to skip this 'pleasing' tone curve, just set Brightness, Contrast and Shadows to zero

                              Thanks, but I still don't want to create linear, nor perceptually linear images. ("Phantom reading" seems to be rampant on these forums.)

                              What I miss is a "cumulated adjustment curve", showing the cumulated effect of the individual adjustments. These adjustments are widely overlapping (building on each other), one can mitigate or eliminate the partial effect of one setting by another. As there is no such feature, I was hoping to understand the effect of at least the contrast setting.

                              Anyway, I realize that this attempt was futile; I give up.
                              • 12. Re: The meaning of "contrast" in ACR 4.1
                                Thomas Knoll Level 2
                                G Sch:

                                If you want a good "starting point for adjustments", just set the controls to whatever makes the image LOOK good.

                                That is the ONLY thing that matters.

                                Why to you care about the underlying math? If you really want understand this, you need to learn some color theory. At least to the point were you fully understand "gamma encoding", "scene referred" data, and "output referred" data. It is clear that you don't, since you seem to be trying to use scene referred data as a starting point for adjustments, which is an awful idea.

                                If you want scene referred data, set the Brightness to zero, Contrast to zero, Shadows to zero, Tone Curve to "Linear". In most cases this results in an ugly, dark, flat image, and would be a terrible "starting point for adjustments". But if you want to shoot yourself in the foot, go ahead.
                                • 13. Re: The meaning of "contrast" in ACR 4.1
                                  Level 1
                                  May be stepping into a hornets' nest here, but:

                                  "If you want a good "starting point for adjustments", just set the controls to whatever makes the image LOOK good."

                                  Wouldn't a good looking image be what you get AFTER adjustments? And if you're interested in experimenting with the various controls to get a better intuitive grasp of their interaction, wouldn't a good place to start from be an image that has had as little adjustments made to it as possible (and which might in fact look terrible to begin with)?

                                  Just seems like G Sch and TK are going are trying to talk while on completely different vectors: TK seems to suggest a workflow of "start with Adobe defaults and make minor changes" while G Sch seems to want a workflow of "remove all Adobe defaults and let me make all of the adjustments from scratch." Both workflows seem reasonable to me. I've used both and have gotten good results from each.

                                  And as to:

                                  "That is the ONLY thing that matters."

                                  I think that experimenting to understand how your tool works in order to be as prepared as possible when confronted by new situations is an equally valid goal.
                                  • 14. Re: The meaning of "contrast" in ACR 4.1
                                    Hudechrome-sd9sPI Level 2
                                    If it looks good, it is good.
                                    • 15. Re: The meaning of "contrast" in ACR 4.1
                                      Panoholic Level 2
                                      >experimenting to understand how your tool works in order to be as prepared as possible when confronted by new situations is an equally valid goal

                                      ACR offers an abundance of controls, which are working additive. The chance to win on lottery is higher than to guess the cumulated effect of recovery, fill light, brightness, contrast, parametric curve adjustment and point curve adjustment.

                                      Anyway, I found a pragmatic way to analyze the individual and combined effect of these controls. I created an image with graduation from black to white, with a flat histogram in sRGB and one in AdobeRGB, in 16-bit TIFF. (I don't know how to achieve this in PS, I managed it in another image editor.) Opening this in ACR, all controls set to zero etc., starting with the flat histogram, the effect of one or more adjustments becomes visible not only on the image, but on the histogram as well. The histogram is quite course, but it is good enough to explain, what is happening.
                                      • 16. Re: The meaning of "contrast" in ACR 4.1
                                        Terminology is so important to a good conversation. This potentially interesting thread got stuck over some terms.

                                        I think the original questioner was confiding he was a bit overwhelmed by the complexity of the controls and seeking a baseline.

                                        The response use what looks good not a suitable baseline in and of itself. There are plenty of ways to arrange controls which look great in one type of image while being pathological wrong in another. Where a different arrangement could provide acceptable results for both.

                                        If you understand the implementation of the control (not just it's immediate effect) then you can better handle the boundary cases. This is the science and art of the use of all tools.

                                        I think TKs response about being clear about the difference between "scene" vs "output" referred image data is important to understanding this conversation. Once the original question is reframed in such terms, maybe it will be easier to answer?

                                        Here is a lucid treatment of these terms by Karl Lang

                                        http://www.adobe.com/designcenter/dialogbox/karllang/pscs3_rendering_image.pdf
                                        --------
                                        From Pages 7 & 8
                                        "If you send scene-referred data directly to a printer, or even a high quality computer display [output], you will not be very happy with the result. Even using the lightness-mapped data will provide a flat, lifeless image. These media dont have the dynamic range thats encoded in the original scene; in fact, far from it. "
                                        [...]
                                        "In order to create a photographic image that evokes some resemblance to the original scene, we need to compress some parts of the tone scale and stretch others. We may also use clipping on at least one end. Clearly this process is both subjective and image-dependent. Theres no universal right way, no formula or curve that will work for every image."
                                        --------

                                        NOTE that gamma coding as implemented in personal computers and digital photography is not about creating pleasing tonality and its design is not intended for tonal effects in a properly color managed system. I think the original poster gets this but maybe not everyone? Gamma is about other concerns: historical trends in display technology and associated costs of input to output transformations (CRTs and half-tone printers), and about making the most of the representation of image data (limited bits per pixel).

                                        The wonderful and tricky thing about understanding this stuff is noticing how designers overcome limits of one part of a system with some clever aspect of design of another part, towards the successful balance for some purpose.

                                        It looked like this thread was gonna reveal something about using ACR /Photoshop in this regard, but then it got stuck!
                                        • 17. Re: The meaning of "contrast" in ACR 4.1
                                          Hudechrome-sd9sPI Level 2
                                          Creating a photograph to evoke some resemblance to the original is but one aspect of producing a photograph. I try to produce photographs that reflect what I felt at that moment. Therefore, I use all the controls available to me, from the exposure onwards, and I draw on my 45+ years of darkroom work to guide my choices at the digital level.

                                          It is instructive to examine Adams "Moonrise" image, with the knowledge that he underexposed the film and resorted to intensification to bring that feeling to the final image. I suspect that it took some years to get there also, but I don't know for sure.

                                          For my clients, however, they do want an image that has some resemblance to the original, so I have to perform at that level as well. There is no good single pathway to that end either. Nonetheless, my most recent review of ongoing photo work with a client resulted in the remark of how well I captured the sense of hot light (summer light in the Northwest can be deadly), something I never mastered as well in all my years in film. How I got there, well, it took more than one visit to ACR to get there. I tweaked until it looked good.

                                          Stuckness lies with the person, not the thread. When you understand that everything you learn about photography are but points of departure for your own work, you are on the way to unstuckness.
                                          • 18. Re: The meaning of "contrast" in ACR 4.1
                                            > If you want to see a linear mapping of your image, set the Exposure & Blacks (4.x) to zero, set Brightness tp -150 & Contrast to -50 and set the point curve to linear..
                                            ( Jeff Schewe, "The meaning of "contrast" in ACR 4.1" #4, 20 Jul 2007 6:46 pm)

                                            > If you want scene referred data, set the Brightness to zero, Contrast to zero, Shadows to zero, Tone Curve to "Linear".
                                            ( Thomas Knoll, "The meaning of "contrast" in ACR 4.1" #12, 23 Jul 2007 10:09 am)

                                            Ahem. So which one it is? :-)

                                            Kaa
                                            • 19. Re: The meaning of "contrast" in ACR 4.1
                                              Panoholic Level 2
                                              > So which one it is?

                                              I described the way I went to find out the effect of the controls, in #15. The result shows clearly, that contrast = 0 and brightness = 0 are the "neutral" setting.

                                              It's the best to repeat the test for yourself, in order to see the effect of several adjustments, in combination as well; some of them can be surprizing, for example Recovery with values over 50.
                                              • 20. Re: The meaning of "contrast" in ACR 4.1
                                                olgol

                                                I came to this thread wondering - if I wanted to edit my RAW photo in Photoshop - what settings would I make in ACR that would be the least destructive to the original file. Setting the blackpoint to zero, Sharpening to zero, Exposure to the right, etc all make sense as they alter the underlying data.  Brightness and contrast are the only two weird ones.  The final reply of Panoholic 19 makes the most sense to the one without the vocablulary.  Thanks!