19 Replies Latest reply on Jan 20, 2010 9:06 AM by Romain_Th

    What is the purpose of the DNG baseline exposure?

    Romain_Th

      I don't think the DNG specifications are very explicites. What kind of consistency does it try to achieve ?

      • Place the 18% grey as midtone for every camera ? I don't think so cause my D700 places it way above.
      • Keep the same amount of RAW headroom ?
      • Get the same white-point as the camera maker rendering ? I think it behaves like that for my D700 but not really for my Sony A700.
      • Get the same white-point for every camera (regardless of the metered scene) ?

      How is it calculated ?

       

      Thank a lot for the information. I'm not trying to blame Adobe for overexposing or anything else like that. I'm sure there is a reason that tag is here, I simply would like to kno why.

       

      Romain

        • 1. Re: What is the purpose of the DNG baseline exposure?
          Bill_Janes Level 2

          Romain_Th wrote:

           

          I don't think the DNG specifications are very explicites. What kind of consistency does it try to achieve ?

          • Place the 18% grey as midtone for every camera ? I don't think so cause my D700 places it way above.
          • Keep the same amount of RAW headroom ?
          • Get the same white-point as the camera maker rendering ? I think it behaves like that for my D700 but not really for my Sony A700.
          • Get the same white-point for every camera (regardless of the metered scene) ?

          How is it calculated ?

           

          Thank a lot for the information. I'm not trying to blame Adobe for overexposing or anything else like that. I'm sure there is a reason that tag is here, I simply would like to kno why.

           

          Romain

          Romain,

           

          The rationale for using the baseline exposure is explained in the DNG specification.

           

          http://www.adobe.com/products/dng/pdfs/dng_spec.pdf

           

          The newer Nikon cameras allow about 0.5 EV of highlight headroom, so their effective ISO needed for 18% saturation of the sensor is less than the nominal ISO. If you expose according to the light meter reading you get 12% saturation as shown below in Rawnalize for the D3, which behaves similarly to the D700. The 12 bit raw value is 497 and the saturation is 497/4095 = 12.1%. The corresponding 8 bit gamma 2.2 pixel value is 99.8

           

          Rawnalize001.png

          The D70 exposure is closer to nominal as shown in a study by Bill Claff:

           

          http://home.comcast.net/~nikond70/Investigations/Sensor_Characteristics.htm

           

          The other factor in the pixel value that you would obtain in the rendered sRGB file is the tone curve used by the camera JPEG engine or the raw converter. Using the Standard Picture control for my D3 gives a gray card pixel value of 150. ACR with the Camera Standard profile and default settings gives a value of 163; correcting for the baseline offset of +0.5 EV by using -0.5 EV exposure in ACR gives an sRGB pixel value of 136. If you use the Adobe Standard profile with a linear tone curve (sliders on main tab at zero and a linear point curve) and the exposure offset of -0.5 EV, you get an Adobe RGB value of 99, very close to the gamma 2.2 rendered raw value with a linear tone curve as in Rawnalize. The Camera standard profile with the same settings gives a 91.

           

          ACR_scrCap.png

           

          So the baseline offset for this camera seems to give a pixel value in the gamma 2.2 rendered pixel value that you would get by applying a linear tone curve to the raw file (aside from the gamma encoding). Personally, I would  prefer no baseline offset, but one can get the desired result by setting the ACR default exposure to -0.5 EV. If you convert your raw file to DNG, the BaselineExposure value is in the EXIF and can be read with any EXIF editor.

          • 2. Re: What is the purpose of the DNG baseline exposure?
            Romain_Th Level 1

            Thanks a lot Bill_Janes, for that explanation!

             

            What do you mean by "The newer Nikon cameras allow about 0.5 EV of highlight headroom"? That the camera JPEGs are clipping 0.5 EV less than the sensor saturation value?

             

            What about negative baseline offsets (ex D200)? What data is pulled back?

             

            Don't you think the default tone curve is way too bright for a corrected exposure? (and better for a 12% metering)

             

            And my last question: why can I pull back almost 2 stops of highlights without any loss whereas I am supposed to have only 0.5 EV ?

             

            Thanks a lot for your help.

            • 3. Re: What is the purpose of the DNG baseline exposure?
              Panoholic Level 2
              function(){return A.apply(null,[this].concat($A(arguments)))}

              Bill_Janes wrote:

               

              The corresponding 8 bit gamma 2.2 pixel value is 99.8

              ...

              you get an Adobe RGB value of 99, very close to the gamma 2.2 rendered raw value with a linear tone curve as in Rawnalize


              Bill,

               

              this has nothing to do with the subject; it is only for your understanding: the currently selected mapping is sRGB, which has a gamma = 2.4 (after a small linear ramp). Gamma = 2.2 is with Adobe 1998 RGB.

               

               

              Gabor

              • 4. Re: What is the purpose of the DNG baseline exposure?
                Bill_Janes Level 2

                Gabor,

                 

                Yes, I forgot to take that into account. However, for midgray (L*= 50), there is not much difference between the sRGB value and that for a straight gamma 2.2 curve. For a gamma of 2.2, L* = 50 corresponds to a pixel value of 118.19, whereas it is 118.92 in sRGB.

                 

                Although Adobe RGB has no linear segment in the defintion, the implementation allows a linear segment and this is used in Adobe Photoshop.

                 

                http://forums.adobe.com/message/1657509#1657509

                • 5. Re: What is the purpose of the DNG baseline exposure?
                  Bill_Janes Level 2

                  Romain_Th wrote:

                   

                  Thanks a lot Bill_Janes, for that explanation!

                   

                  What do you mean by "The newer Nikon cameras allow about 0.5 EV of highlight headroom"? That the camera JPEGs are clipping 0.5 EV less than the sensor saturation value?

                   

                  What about negative baseline offsets (ex D200)? What data is pulled back?

                   

                  Don't you think the default tone curve is way too bright for a corrected exposure? (and better for a 12% metering)

                   

                  And my last question: why can I pull back almost 2 stops of highlights without any loss whereas I am supposed to have only 0.5 EV ?

                   

                  Thanks a lot for your help.

                   

                  To get the ISO calibration for various Nikon cameras look at Bill Claff's Sensor Analysis page:

                  http://home.comcast.net/~nikond70/Investigations/Sensor_Characteristics.htm

                   

                  The newer Nikons such as the D3 and the D300 have their effective ISOs less than the nominal value. For example, the D3 has a base ISO nominal value of 200, whereas the effective ISO to give 18% saturation in the raw file is about 138. The D70 and D200 are near nominal. For example, the D3 underexposes by 0.5 EV and uses a hot tone curve. The BaselineExposure for the D3 is +0.5 EV and the values for the D70 and D200 are -0.75 and -0.5 respectively.The purpose of the BaselineExposure is to get similar results for the rendered raw images when different cameras are used.

                   

                  Since the D3 gives 12% saturation for the metered area rather than 18%, this means that all values for the histograms are shifted to the left by 0.5 EV. If the effective exposure was for 18% and the highlights were at clipping in the raw file, then they would be 0.5 EV short of clipping for the 12% calibration. This would amount to 0.5 EV of highlight headroom.

                   

                  IMHO the D3 tone curve is too bright, but this allows extra protection for the highlights and allows conisderably more highlight recovery in ACR when the values appear clipped in the rendered file but not in the raw file. The same would apply to the D700 which behaves very much like the D3.

                   

                  Highlight recovery in ACR for clipped raw files requires intact channels in the raw file. For example, here is a raw histogram for the D3 with the green channel at clipping:

                  RawnalizeHisto.png

                  The red channel is about 2/3rds stop below clipping and the blue channel about a half stop short of clipping in the histogram. For white balance the red channel is multiplied by 1.825 and the blue channel by 1.645 (I calculate the increase in f/stops to be 0.87 and 0.72 f/stops respectively). If you increased exposure by 0.5 EV, the green channel would be clipped, but there would still be intact data in the blue and red channels, and highlight recovery would be possible. If you increased exposure by more than 0.87 stops, all channels would be clipped and highlight recovery would not be possible. If you can recover 2 stops in ACR, this means that the tone curve was too hot and the histogram indicated clipping in the rendered file while the raw file was in fact intact.

                  • 6. Re: What is the purpose of the DNG baseline exposure?
                    Romain_Th Level 1

                    Actually, by default tone curve, I meant the standard ACR one found in the Adobe Standard or in generated DNG profils, not the D700 Camera Standard.

                    This Adobe Standard curve (which is the result of Brightness +50 and Contrast +25) is very much like the Canon 5D II standard curve and also quite similar to the 1Ds series and 5D I.

                     

                    I am asking all this because using a Canon 5D II spot meter seems to place the metered luminance in the center of the histogram (as are most of the recorded pictures) with the adobe standard curve or the canon camera standard one. I know it is the result of the curve boosting the 18% saturation (or corrected saturation) at the middle.

                     

                    But with the D700 (=D3), the mettered luminance is way on the right :

                    Capture d’écran 2010-01-08 à 08.30.44.png

                    And so are most of rendered images using the Adobe Standard tone curve. I find this to bright, but don't find canon 5D or 5D II too bright.

                     

                    So I am asking what is going on ? Since they should render approximatively the same...

                     

                    PS: Putting the "grey point" in the middle is very useful because adjusting the contrast will not change its brightness.

                    • 7. Re: What is the purpose of the DNG baseline exposure?
                      MadManChan2000 Adobe Employee

                      At present, all of the Adobe Standard profiles use the same tone curve. (More specifically, they do not contain a tone curve, which means it's up to the host application - Camera Raw in this case - to provide the tone curve. And that default curve for Camera Raw is the Brightness +50 / Contrast + 25 / Point Curve = Medium Contrast combo.)

                       

                      Cameras do meter differently but that's not something we can do much about because cameras will sometimes vary their metering depending on shooting mode. (For example, I've seen models that will change how they meter depending on whether you're in Av mode or Manual mode, even if you're using spot meter.)

                       

                      Eric

                      • 8. Re: What is the purpose of the DNG baseline exposure?
                        Bill_Janes Level 2

                        Just to sumarize my results for a shot of a gray card with the D3:

                         

                        ExceGrayCard.gif

                        • 9. Re: What is the purpose of the DNG baseline exposure?
                          Romain_Th Level 1

                          Thanks Eric, I assume the DNG profile curve (when defined) override ACR Brightness/Contrast/Point curve. 50/25/default is used like 0/0/flat (not seen by the user).

                           

                          Bill, I asked a file on a Canon forum to get the same type of results with a 5DII. We will see the differences. Thanks.

                           

                          By the way, I am not sure about the "Linear" lines : putting the sliders at 0 (and point curve to flat) for custom tone curves would apply an inverted ACR default tone curve to them. Since they are not the same, the result would not be exactly linear.

                          • 10. Re: What is the purpose of the DNG baseline exposure?
                            Bill_Janes Level 2

                            Romain_Th wrote:

                             

                            By the way, I am not sure about the "Linear" lines : putting the sliders at 0 (and point curve to flat) for custom tone curves would apply an inverted ACR default tone curve to them. Since they are not the same, the result would not be exactly linear.

                            I don't think your theory about an "inverted tone curve" is correct. One can determine the tone curve experimentally with a Stouffer step wedge and Imatest. Perhaps Eric can comment.

                             

                            Image rendered by ACR into Adobe RGB with a linear tone curve. The shadows are deviate from linearity becuase of the linear segment in the shadows. The values are still gamma encoded and not truly linear.

                             

                            D3_0007_acr_lin_aRGB_Step_3.png

                             

                             

                            sRGB uses a greater shadow deviation:

                             

                            D3_0007_ACR_lin_sRGB_Step_3.png

                             

                            Computed gamma 2.2 vs sRGB encoding to demonstrate the linar segment in sRGB

                             

                            sRGBvsG2_2.gif

                            • 11. Re: What is the purpose of the DNG baseline exposure?
                              Romain_Th Level 1

                               

                              Here are the results of a 1DMKIII with Adobe Standard:

                              Defaults: 130
                              Defaults -0.35EV: 110
                              Linear: 96
                              Linear -0.35 EV: 85
                              As you can see, there are not the same as the D3/D700 and allow a good use of the standard tone curve (exposed scenes are well centered in the histogram).
                              In addition, the baseline offset doesn't put the metered grey at 18% but rather at 12%. So if you add 3EV to the exposure you get the histogram spike just at clipping (more like the ANSI standards I think) instead as 2.5EV for the D3 (more like the ISO standard I think)
                              As a result, that BaselineExposure does not permit to use the same tone curve for every camera to get the same result, and I still don't know what is the point of it (and how it is calculated)...

                               

                              • 12. Re: What is the purpose of the DNG baseline exposure?
                                Bill_Janes Level 2

                                Romain_Th wrote:

                                 

                                 

                                 

                                Here are the results of a 1DMKIII with Adobe Standard:

                                Defaults: 130
                                Defaults -0.35EV: 110
                                Linear: 96
                                Linear -0.35 EV: 85
                                As you can see, there are not the same as the D3/D700 and allow a good use of the standard tone curve (exposed scenes are well centered in the histogram).
                                In addition, the baseline offset doesn't put the metered grey at 18% but rather at 12%. So if you add 3EV to the exposure you get the histogram spike just at clipping (more like the ANSI standards I think) instead as 2.5EV for the D3 (more like the ISO standard I think)
                                As a result, that BaselineExposure does not permit to use the same tone curve for every camera to get the same result, and I still don't know what is the point of it (and how it is calculated)...

                                 

                                Romain,

                                 

                                Thanks for your input. For discussion, you might want to look at Doug Kerr's article on Exposure Calibration and the Wikipedia article on film speed, which dicusses the ISO spec for calibration of digital sensors.

                                 

                                http://doug.kerr.home.att.net/pumpkin/Exposure_Calibration.pdf

                                http://en.wikipedia.org/wiki/Film_speed

                                 

                                Doug discusses ISO 2721 (light meter calibration) and ISO 12232 (digital sensor calibration). In essence, the 2721 standard is calibrated for an effective reflectance of 12.8%. Nikon and Canon both seem to use this standard; otherwise the light meter readings taken by the camera and by a hand held light meter would give different results. The 12232 saturation standard states that exposure according to a standard light meter reading should give a sensor saturation of 12.7%, which is a half stop below the mid gray value of 18%.

                                 

                                Appendix B of Doug's article quotes a Canon USA source stating that a uniform target exposed according to the camera meter reading should give an Adobe RGB value of 55% of saturation, which would be 140. This corresponds to a relative luminance of 17.3% [((100-55)/100)^2.2], which means the sensor saturation is about the same value if the camera did make significant alterations from a gamma 2.2 tone curve. This means that Canon rates their sensor at a lower ISO than would be derived from the 12232 standard so that the sensor saturation is 17.3% rather than the ISO standard of 12.8%. In essence, Canon is eliminating the 0.5 EV fudge factor in the ISO sensor standard mentioned above. This is likely what most photographers would want.

                                 

                                On the other hand, the D3 seems to give around 12% saturation with the light meter reading in compliance with the ISO standard, and then applies a hot tone curve.

                                 

                                As you state, the baseline offset does not seem to give the same results with various cameras, and hopefully Eric Chan or some other Adobe expert will explain things for us.

                                • 13. Re: What is the purpose of the DNG baseline exposure?
                                  Romain_Th Level 1

                                  Thanks for that!

                                   

                                  I understand now why my D700 and my Sony A700 needed -0.5EV to work well with the standard curve (witch is very much like a canon standard curve).

                                   

                                  I guess most of the DSLR will behave the same in ACR but not the Canon ones. So the standard curve is way more acceptable to be used with Canon than other cameras (or you have to put -0.5EV to their exposures).

                                   

                                  Maybe there is something to be done...

                                  • 14. Re: What is the purpose of the DNG baseline exposure?
                                    Romain_Th Level 1

                                    Actually, If you check out these pages (at the exposure times sheets) :

                                    http://www.imaging-resource.com/PRODS/D700/D700A5.HTM

                                    http://www.imaging-resource.com/PRODS/E5D2/E5D2IMAGING.HTM

                                    You will see that under controlled lightning, the 5dMKII is exposing twice the time for the same scene... which is quite the opposite of what have been said.

                                     

                                    I'm beginning to wonder if the BaselineExposure tag has been designed to balance shadow-boost/recovery the same for every camera rather that constant exposure with default metering, forcing you to underexpose by 0.5EV.

                                    Thus I am asking myself if I should better reduce 0.5 exposure in the camera or in LR.

                                    • 15. Re: What is the purpose of the DNG baseline exposure?
                                      Hudechrome-sd9sPI Level 2

                                      The D90 compared to the D80 Nikons is definitely overexposing, (among other differences). I have the camera set to 1/3 stop underexpose, and I get much better results overall. Pushing everything to the right has it's own difficulties.

                                       

                                      BTW, when running into a large dynamic range that has you wondering :"what exposure?" it's wise to go to the lowest ISO useable, as that gives the greatest dynamic range to handle matters. Usually, that's outdoors in bright sun/deep shade and a low ISO, especially coupled with VR, is easy to accommodate.

                                      • 16. Re: What is the purpose of the DNG baseline exposure?
                                        Panoholic Level 2

                                        1. According to the DNG converter, the difference between the "effective exposure" of the Nikon D700 and the Canon 5DMkII is only 1/10 EV (the D700 image needs to be pushed by 0.5 EV, the 5D2 image only by 0.4 EV).

                                         

                                        2. I don't really understand the question if you should reduce the exposure when shooting or when processing the image. If the "normal" i.e. non-adjusted exposure causes clipping in the raw channels, then you have to reduce the exposure when shooting if you want to avoid the clipping; otherwise reduce it in raw processing, if you find that necessary, though "recovery" is often better. Keep in eyes, that the "Brightness" adjustment is just like "Exposure", except in the highlights. Brightness=+50 is Exposure=+1 EV.

                                         

                                        Gabor

                                        • 17. Re: What is the purpose of the DNG baseline exposure?
                                          Romain_Th Level 1

                                          Gabor,

                                           

                                          1. It has been said earlier on this thread that the offset is is based on the camera metering (to place the metered area to 18% sat). Do you think it could have been based on the actual exposure ? This is completely different.

                                          As for the actual exposure of both camera, check out this:

                                          http://www.dxomark.com/index.php/eng/Image-Quality-Database/Compare-cameras/(appareil1)/29 6|0/(appareil2)/305|0/(onglet)/0/(brand)/Nikon/(brand2)/Canon (ISO sensitivity tab)

                                          You can see that the D700/D3 is -0,3EV at every real ISO but the 5DII starts at -0,4 at ISO200 until -0,8 at ISO6400. DXO marks are very serious sensor tests. As for the scene on Imaging Resources, I agree the differences could come from metering and curves.

                                          Maybe there are different teams that calibrate cameras... The Canon team calibrate to get real ISO, and the Nikon one to get 18% metering...

                                           

                                          2. In fact I meant for the every day easy shooting with matrix metering where I don't want to fiddle with the exposure slider in ACR, since ACR eat 1/2EV the camera supposed to have when it metered the scene, should I use default 1/2EV underexposure in the camera and get a constant 1/2EV headroom in the files if needed or should I cancel that "eating" by applying -1/2EV in ACR as default ?

                                           

                                          Romain

                                          • 18. Re: What is the purpose of the DNG baseline exposure?
                                            Panoholic Level 2
                                            function(){return A.apply(null,[this].concat($A(arguments)))}

                                            Romain:

                                             

                                            You can see that the D700/D3 is -0,3EV at every real ISO but the 5DII starts at -0,4 at ISO200 until -0,8 at ISO6400. DXO marks are very serious sensor tests.


                                            DNG files show, what Adobe "thinks" of the native image, and the BaselineAdjustment parameter does not change with the ISOs (except with ISO 50, which is in fact an overexposed ISO 100). Previously I have not seen DXO's measurements related to the 5D2's ISO values; now I see two contradicting statements. I have no way to verify their measurement myself.

                                             

                                            Still, there is one thing I can say with certainty, that DXO is wrong about: ISO 3200, the highest "real" ISO of the 5D2, is stated as "measured ISO" 2133. 3200 divided by 2133 is 1.5... ISO 6400 is stated to be 3990, this yields 1.6; 12800 is stated to be 7550, yielding 1.695, the same as with 25600.

                                             

                                            However, 6400, 12800 and 25600 are fake ISOs; the 5D2 creates the pixel values by multiplying the values gained by the highest real ISO. Thus the fake ISOs can not have a different "sensitivity" than the last real ISO's "sensitivity". This makes me doubting the validity of the DXO values.

                                             

                                            (I don't have any personal interest in this issue.)

                                             

                                            Gabor

                                            • 19. Re: What is the purpose of the DNG baseline exposure?
                                              Romain_Th Level 1

                                              Yes, you are perfectly right, I didn't notice that. Maybe the digital gain on fake ISO is not linear to force the user to expose more and get little less noise... but I doubt.