23 Replies Latest reply on Jan 26, 2011 2:42 PM by paul_48

    Question on Recovery in ACR.

    paul_48 Level 1

      My camera is set to Uni-WB and I ETTR. For the vast majority of images, the ACR histogram shows highlights set to the very right side (as expected) with Exposure set at 0. (Of course this is after I've set WB). Occasionally I have an image that shows that highlights are clipped in all 3 channels as indicated by white showing in the triangle (probably due to my getting too close to the right when exposing).  I understand (per a Bruce Fraser post) that when all channels are clipped nothing can be done to recover the highlights.  (And my attempts to recover with either Exposure or Recovery sliders visually bear this out).  What is bothering me is that when I move Recovery to the right, the white triangle (and the clipping warning when opt is held) both indicate that there is not longer any clipping. It seems to me that these indicators should continue to stay at white (and red) rather than indicate that clipping is not longer present.  Comments?

        • 1. Re: Question on Recovery in ACR.
          RASouthworth Level 3

          Clipping is a function of the ACR  chosen rendering space, by default sRGB.  Often not every channel in the raw data is clipped even if so indicated because of the limitations of sRGB, and using recovery or exposure will bring the channels into range.  Choose ProPhoto as the rendering space to obtain a truer look at channel clipping.

           

          Richard Southworth

          • 2. Re: Question on Recovery in ACR.
            Noel Carboni Level 7

            RASouthworth has hit the nail on the head here I think...

             

            With ACR 6.3 and using ProPhoto RGB and starting at 0 EV I have not been able to duplicate what you're seeing with the Recovery slider causing a change in the highlight clipping indicator.  I just ran through a number of images with even marginal partial overexposure and in every case a white highlight clipping indicator stayed white.  Said another way, when the clipping indicator was white I couldn't get it to change at all by using either the Exposure or Recovery sliders.  I could get it to change if it was colored in a couple of cases (e.g., sunset clouds with bright red colors).

             

            Moral:  Convert to ProPhoto RGB.

             

            -Noel

            • 3. Re: Question on Recovery in ACR.
              paul_48 Level 1

              I should have mentioned - I am in ProPhoto, 16bit.

              I'm not sure how I can post image as it is 21MB (raw) from a 5DII.  I'll see if I can find a way.

              • 4. Re: Question on Recovery in ACR.
                Noel Carboni Level 7

                Interesting.  Maybe it could have something to do with the profile you're using?  I'm testing with the Camera Standard profile.

                 

                I would like to see your file (as, I'm sure, would the Adobe guys).  YouSendIt.com will host large files - send it to yourself then post the link here.

                 

                I actually did some more testing, and while I can't get the clipping indicator to go away with just the Recovery slider, I can make it go off with huge negative Exposure settings on a test image of mine that has only marginal overexposure.  This is that image:

                 

                http://Noel.ProDigitalSoftware.com/temp/IMG_3946.zip

                 

                I can't explain why, unless perhaps math error resulting in a small value change may be playing up, that this should occur.

                 

                -Noel

                • 5. Re: Question on Recovery in ACR.
                  paul_48 Level 1

                  I have saved as a jpg at 1.5MB and it shows the same action: Although highlights show as blown, when Recovery is moved past about 60 the triangle is not longer white.  I'll admit that I'm not sure what saving it as jpg (directly from ACR) would do - it's something I've never had a reason to do. I've put the jpeg on: http://osgood.smugmug.com/Other/acrquestion/15556089_3oTGV#1165322905_NqMjQ-A-LB

                  • 6. Re: Question on Recovery in ACR.
                    Noel Carboni Level 7

                    Curiouser and curiouser...  The small part of a roof section in my image that's actually overexposed and unrecoverable, when opened with minimal Exposure and maximal Recovery settings, then further manipulated to darken it further shows indeed that there is no detail, but a very interesting radial gradient of values appears that are less than pure white.

                     

                    SmallRoofSection.jpg

                     

                    This gradient is so smooth that I have to wonder whether it's a processing side effect perhaps of the logic behind the Recovery slider that blends recovered sections with darker surroundings.

                     

                    -Noel

                    • 7. Re: Question on Recovery in ACR.
                      RASouthworth Level 3

                      I found a raw image that did the same as yours, i.e. showed 255-255-255 (ProPhoto) in a particular area but still could be "tamed" with recovery and exposure.  I opened in Photoshop, changed the info pointer to 16 bit, and found some headroom, i.e. pixel channels measured less than 32768 even though they were 255 in 8 bit.  Unfortunately one can't do the 16 bit measurement directly in ACR, so the round-off to 8 bit may indicate clipping when not actually occurring.

                       

                      Richard Southworth

                      • 8. Re: Question on Recovery in ACR.
                        paul_48 Level 1

                        Unfortunately I'm not sure how I can send it.  My email restricts sent msgs to 20MB and even zipped it is larger.

                        • 9. Re: Question on Recovery in ACR.
                          paul_48 Level 1

                          I've opened the original .CRE file (not the jpg) in PS5 and info shows 32768 in all 3 channels.

                          • 10. Re: Question on Recovery in ACR.
                            Noel Carboni Level 7

                            No need to send it.  As you can see above I've found an image that behaves the same.

                             

                            I don't think it's 8 vs. 15 bit...  I think there's an additional level of "magic" in the processing pipeline (visible as the gradient I noted above) that's causing the effect.

                             

                            No one really knows what "Recovery" really means in its current implementation, and I doubt Adobe is going to share every intimate detail.

                             

                            -Noel

                            1 person found this helpful
                            • 11. Re: Question on Recovery in ACR.
                              paul_48 Level 1

                              Thanks - It'll be interesting to see if we get further explanation.  This happens with about 5% of my images.  Since I'm new to digital (previously med format film) I'll now back off a bit from exposing to the very right.

                              • 12. Re: Question on Recovery in ACR.
                                Noel Carboni Level 7

                                Fully understanding the merits of ETTR, and knowing the downsides, I have finally decided that exposing "properly" per the camera's settings and histogram is the best philosophy for me.

                                 

                                Note also that there are several downsides, such as color shifts or loss of saturation, to ETTR that we haven't discussed in this thread.

                                 

                                The extra "headroom" provided by my "EPPC" (Expose Properly Per the Camera) strategy - with my camera and the way I shoot - has saved an otherwise troublesome shot for me on more than one occasion, and my camera simply doesn't deliver so much noise that I can't stand it if I actually even have to lighten an image.

                                 

                                -Noel

                                • 13. Re: Question on Recovery in ACR.
                                  Jeff Schewe Level 5

                                  paul_48 wrote:

                                   

                                  Thanks - It'll be interesting to see if we get further explanation.

                                   

                                  Well, you might be surprised just how much data can be in the brightest stop of your capture. I often find that clipping that shows in all three channels can often be eliminated often by only slight -exposure or adding recovery. It often depends on just how far the clipping has gone. Totally blown RGB will be forever clipped but often what the clipping indicator is showing is a lot of clumping at or near the clip point.

                                   

                                  ETTR is useful for getting an optimal signal to noise ratio but there's a limit how far you should take it. Over exposed is over exposed. I've written a bit about it in this article Un-debunking ETTR. Also note the second image that shows just how much data can be in the far right highlight of an image...there's a difference between clipping and clumping.

                                  • 14. Re: Question on Recovery in ACR.
                                    paul_48 Level 1

                                    Jeff  - Thanks for the insight and the article "Un-debunking ETTR".  When I converted from med-format film to digital a few months ago, I learned about ETTR and noticed that the 5DII's histogram did not correspond to that in ACR.  This led me to Uni-WB which does give me a much more reliable histogram. I have been exposing virtually all images with ETTR, even those requiring less than the full dynamic range of the camera. After setting the WB they typically need no adjustment in Exposure, just a little Recovery.  If I've read your article correctly, my technique is correct. Please inform me if I should not  ETTR for normal contrast scenes. I'm still not sure why the white triangle no longer shows blown highlights when there are pixels at 32768 for all 3 channels. If I'm exposing optimally then the reason is only academic and I not relevant to my shooting.

                                     

                                    Thanks,

                                    Paul 

                                    • 15. Re: Question on Recovery in ACR.
                                      Bill_Janes Level 2

                                      Jeff Schewe wrote:

                                       

                                      paul_48 wrote:

                                       

                                      Thanks - It'll be interesting to see if we get further explanation.

                                       

                                      Well, you might be surprised just how much data can be in the brightest stop of your capture. I often find that clipping that shows in all three channels can often be eliminated often by only slight -exposure or adding recovery. It often depends on just how far the clipping has gone. Totally blown RGB will be forever clipped but often what the clipping indicator is showing is a lot of clumping at or near the clip point.

                                      The essential thing is to look at the raw file. Due to the white balance multiplers, the red and blue channels may be clipped in ACR, but they may be intact in the raw file. Rawnalize is a good tool for this if your camera is supported. DCRaw or Iris can also be used for this purpose. I would invite readers to look at the White Balance section of Guillermo's DCRaw tutorial. The same principles apply to ACR.

                                       

                                      http://www.guillermoluijk.com/tutorial/dcraw/index_en.htm

                                       

                                      In addition, clipping in ACR may be due to the BaselineExposure value that ACR uses. The raw file channels may be totally intact. This happens with the Nikon D3 where ACR uses a BaselineOffset of +0.5 EV.

                                      • 16. Re: Question on Recovery in ACR.
                                        Vit Novak Level 3

                                        I presume that recovery tool essentialy recovers informations that were not clipped in raw, but became clipped after whitebalancing and exposure correction, by compressing highlight tones prior to that. However, there is obviously some additional logic built into recovery algorithm, that is trying to guess some information that are actually clipped in raw file, because this algorithm sometimes leads to some strange hue shifts in highlights when you apply recovery slider past the highlight headroom (for 400D, where BaselineExposure is 0.25, this occurs at recovery slider position 6 or 7).

                                        • 17. Re: Question on Recovery in ACR.
                                          Bill_Janes Level 2

                                          Vit Novak wrote

                                           

                                          I presume that recovery tool essentialy recovers informations that were not clipped in raw, but became clipped after whitebalancing and exposure correction, by compressing highlight tones prior to that. However, there is obviously some additional logic built into recovery algorithm, that is trying to guess some information that are actually clipped in raw file, because this algorithm sometimes leads to some strange hue shifts in highlights when you apply recovery slider past the highlight headroom (for 400D, where BaselineExposure is 0.25, this occurs at recovery slider position 6 or 7).

                                          Vit, Your assumption is correct. If all three channels are blown, there is nothing to recover. ACR reconstructs the blown highlights to neutral, which is not unreasonable, since highlights usually have low chroma. DCRaw has an option to reconstruct highlights from the adjacent non-blown areas as demonstrated in Guillermo's tutorial. Look at the woman's face in Figure 9.

                                           

                                          http://www.guillermoluijk.com/tutorial/dcraw/index_en.htm

                                          • 18. Re: Question on Recovery in ACR.
                                            Peter DL Level 1

                                            >>I understand (per a Bruce Fraser post) that when all channels are clipped nothing can be done to recover the highlights.  …  What is bothering me is that when I move Recovery to the right, the white triangle (and the clipping warning when opt is held) both indicate that there is not longer any clipping.<<

                                            Baseline Exposure was already mentioned. So when exposure looks clipped, well then,
                                            just drag the Exposure or Recovery slider max. to the left (minus values), and IF white is still 255 white, then it is completely clipped, probably un-recoverable.

                                            Whether blow highlights (single channels) should be reconstructed to "neutral" is a different question. Scene Chroma does not care about exposure, DR restrictions, the Tone Curve applied and gamut limitations.

                                            Peter

                                            --

                                            • 19. Re: Question on Recovery in ACR.
                                              paul_48 Level 1

                                              Thanks Peter. 

                                               

                                              I am aware that I can use Exposure or Recovery (which I prefer since it does not affect the shadows and highlights).  What I'm pointing out is that the indicators should not be trusted.  After reducing the exposure or increasing recovery, it is not sufficient to look at the triangles to know if the clipping is gone.  In addition you need to carefully examine all the highlight areas and see if they include any pixels at 255 in all channels. I'm also wondering if this is a bug and if it would not be better to show the triangle as white whenever there is highlight clipping. 

                                              • 20. Re: Question on Recovery in ACR.
                                                Vit Novak Level 3

                                                Instead of that triangle, you can also use overexposure warning. After turning it on, move recovery slider to the right until blown areas (painted red) dissapear. To make sure what is blown and what is indeed of red color, you can turn it on/off/on/off and see

                                                • 21. Re: Question on Recovery in ACR.
                                                  paul_48 Level 1

                                                  Thanks Vit,

                                                   

                                                  While I have used the overexposed (red) warnings, but until a minute ago, I had not noticed that this warning is more accurate than the triangle.  When using Recovery, the triangle will turn black (I assumed indicating there is no clipping) when the overexposure is showing areas in red.  As I continue to increase Recovery, the red disappears at the same time as the highlight areas drop to less than 255.  So it appears that the overexposure (red) is an accurate warning of when/where there are pixels at 255,255,255.   Helpful - I'll make sure I leave this on in the future.  I'm now under the belief that the color sampler tool showing 255, 255, 255 does not indicate that all channels are clipped. This is because such areas can be tamed via Recovery.

                                                   

                                                  Paul

                                                  • 22. Re: Question on Recovery in ACR.
                                                    Vit Novak Level 3

                                                    It looks like overexposure warning (red color) means that at least one of color channels (R,G or B) has value 255 in selected output color space (and not in actual raw data). Depending on color space (sRGB, Adobe RGB etc) different areas will be painted red, but for that areas, RGB indicator below the histogram will show 255 for at least one channel

                                                    • 23. Re: Question on Recovery in ACR.
                                                      paul_48 Level 1

                                                      I want to thank everyone who answered my post. 

                                                      I was operating on assumptions of what it meant pixels being at 255 in all channels.  From the responses this does not necessarily mean that the highlights are blown.  As Jeff said this may just indicate 'clumping'.   When I have a little more time, I'll look into other raw converters or ways to see the values of the highlight outside of ACR, such as suggested by Bill.

                                                       

                                                      I've marked this post as answered....Paul