18 Replies Latest reply on May 23, 2013 9:56 AM by station_two

    PV2012 locals sometimes don't behave well...

    areohbee Level 5

      Here's one example:

      bad_moon.jpg

      The squiggles are due to local paint with highlights = -30.

       

      Note: global highlight reduction of same, or greater strength, does not result in so much darkness, nor so much desaturation.

      Note2: marked darkening/grayness/desaturation occurs even at much lower values for -highlights, e.g. -5 (I increased value to exagerate the problem).

       

      Here is the raw+xmp, zipped:

       

      http://www.robcole.com/LrForumSupport/20120505RC193027NB4328.O.zip

       

      I have noticed many other less extreme but similarly-flavored cases.

       

      Others have reported similar "unexpected" (to the layman) behaviors too, e.g.

       

      http://forums.adobe.com/thread/1214206

      (see posts about white bed-sheet made "gray-ish" by applying -exposure local, such did not occur when reducing exposure globally)

       

      Rob

        • 2. Re: PV2012 locals sometimes don't behave well...
          areohbee Level 5

          Either way: doesn't matter... - download the raw if you want to see for yourself.

          PS - I sent this raw to Eric Chan a year or so ago, but I never received any acknowledgement.

          • 3. Re: PV2012 locals sometimes don't behave well...
            Vit Novak Level 3

            In your example, it seems that auto mask was used, because only "desturated" parts of the moon are masked, although I'm not able to get similar mask. If I delete the mask and make a new one (without auto mask checked), I don't get that effect. And "desaturated" parts are actually not desaturated, they are only darker

             

            That said, if you move almost every slider to the extreme position, you can expect these things from any software ...

            • 4. Re: PV2012 locals sometimes don't behave well...
              areohbee Level 5

              Vit Novak wrote:

               

              In your example, it seems that auto mask was used, because only "desturated" parts of the moon are masked, although I'm not able to get similar mask.

              Is there some bug in the auto-mask I am not aware of? (I'm about 99.9% sure it was *not* enabled, but does it matter? why?? - it shouldn't, right???).

               

               

               

              Vit Novak wrote:

               

              If I delete the mask and make a new one (without auto mask checked), I don't get that effect.

              Did you see the gray squiggles before deleting the mask (as posted in this thread above)? I get the same effect whether auto mask is checked or not, I mean other than it being a little stippled when auto-mask is checked. That's not how it's working for you?

               

               

               

              Vit Novak wrote:

               

              "desaturated" parts are actually not desaturated, they are only darker

              That's true for me in the redraw too, but not the original even when using the same settings. The original looks gray, not dark orange like the redrawn mask. Anyway, is there an explanation for why original (converted) paint would not look the same as newly applied paint, when using the same settings (and with the same amount of paint applied)?

               

               

               

              Vit Novak wrote:

               

              That said, if you move almost every slider to the extreme position, you can expect these things from any software ...

              I didn't move any sliders to extreme positions, so I don't get your point here. I posted the example with local highlights=-30 so it was blatantly obvious, but the problem is evident at -5, and looks really bad even at -10 (those are hardly extreme values), and the problem occurs when other settings are at light, moderate, and/or strong values.

               

               

              [edit]:

              Have you ever noticed how locals are sometimes different than (what should be roughly equivalent) globals? I mean, I notice it regularly, as do others. And I'm not talking about non-linearity or diminishing returns when a global setting (or stacked local) is near to max... - not talking about that at all..., nor am I talking about increasing local exposure on an already bright area or decreasing local exposure of an already dark area, which has it's own set of problems due to order of processing locals before globals as it seems to).  This (moon shot) is just one example, and there is another example in this thread which references an example in another thread. And when local effects are substantially different than global effect, the local effect is sometimes not as good, which is why some of us will change targeting in midstream: if local looks like hell, then delete it, change globals to target the area previously painted, and redraw everywhere but the area previously painted.

              [/edit]

               

              Rob

               

              Message was edited by: Rob Cole

              • 5. Re: PV2012 locals sometimes don't behave well...
                Vit Novak Level 3

                Yes, I used your raw and xmp file and I see above posted picture

                 

                The fact is - only blown parts of the moon are painted (doesn't matter did you make it with auto select or manually) - you can see that by setting all controls for the painted area to zero and checking show mask box or moving a mouse pointer to the anchor point of selected area. If you check xmp file (it's a text file), you can see several hundred points belonging to the mask defined

                 

                So if you apply negative highlights or exposure, masked areas get darker. It is to be expected, isn't it ?

                 

                As about "extreme" positions of sliders, well, they are a bit extreme to my taste because I never use that much correction with various sliders and that way, but it's just my personal taste. Main reason is the shape of the masked area

                 

                 

                Edit: as about global vs local adjustments - yes, I already noted that effect could be different for the same amount of correction and posted that remark here about 2 or so years ago. Eric responded that it's because local adjustments are performed after global and in limited precision. Things improved since (as he promised), but I believe there are still some differences

                • 6. Re: PV2012 locals sometimes don't behave well...
                  areohbee Level 5

                  Vit Novak wrote:

                   

                  Yes, I used your raw and xmp file and I see above posted picture

                  Well, at least we've that much for common ground .

                   

                   

                  Vit Novak wrote:

                   

                  The fact is - only blown parts of the moon are painted (doesn't matter did you make it with auto select or manually) - you can see that by setting all controls for the painted area to zero and checking show mask box or moving a mouse pointer to the anchor point of selected area.

                  I painted it by eye, after noticing some over-brightness. But I see nothing blown. Here is the histogram after all settings (except camera profile) have been reset:

                   

                  _bad_moon_histo.gif

                  There is no clipping.

                   

                   

                  Vit Novak wrote:

                   

                  If you check xmp file (it's a text file), you can see several hundred points belonging to the mask defined

                  I don't see how this is relevant to the problem.

                   

                   

                  Vit Novak wrote:

                   

                  So if you apply negative highlights or exposure, masked areas get darker. It is to be expected, isn't it ?

                  You seem to be missing the problem:

                  1. initial (painted in previous version) region is gray. you have to delete it and repaint for it to become orange (i.e. the correct color).

                  2. even very small amount of -highlights has extreme effect, relative to global -highlights or (global or) local exposure. so much so that local highlight slider is essentially unusable - looks fine if one uses exposure instead, but I see this as an anomaly.

                   

                   

                   

                  Vit Novak wrote:

                   

                  As about "extreme" positions of sliders, well, they are a bit extreme to my taste because I never use that much correction with various sliders and that way, but it's just my personal taste.

                  As I said in previous post, the problem exists even if you back the sliders off a little, or even if you back the sliders off a lot.

                   

                   

                  Vit Novak wrote:

                   

                  Main reason is the shape of the masked area.

                  Huh? .

                   

                   

                  R

                  • 7. Re: PV2012 locals sometimes don't behave well...
                    Vit Novak Level 3

                    I think I see the reason for some misunderstandings - from xmp file it looks like you are using LR and I'm using ACR, so I don't understand the point about region painted in previous version ...

                     

                    Anyway, what I see in ACR is that you painted only --close to being blown-- parts of the moon and not whole moon (they are actually not blown in raw file, although they are almost blown / highly compressed with sliders on the global page), and those are parts that appear darker if you apply local highlights slider to them. Highlights control work on the similar principle as highlights and shadows in PS and it's not intended to be used on very small areas like you did - it is (as I believe) based on a kind of wide area USM and you need larger area for it to work properly

                     

                    About "extreme" positions I mean global sliders. Highlights are not blown in raw file, but are highly compressed after whitebalancing, usage of tone curve in the profile and color space conversion and further with your exposure slider setting (that's how PV2012 works - it rarely clips highlights, because a kind of soft clipping / compression is used instead of hard clipping in PV2003 at some places in the workflow)

                    • 8. Re: PV2012 locals sometimes don't behave well...
                      areohbee Level 5

                      Vit Novak wrote:

                       

                      I think I see the reason for some misunderstandings - from xmp file it looks like you are using LR and I'm using ACR, so I don't understand the point about region painted in previous version ...

                      I meant previous process version. - I don't imagine there being any difference whether LR or ACR (?)

                       

                       

                      Vit Novak wrote:

                       

                      Anyway, what I see in ACR is that you painted only --close to being blown-- parts of the moon and not whole moon (they are actually not blown in raw file, although they are almost blown / highly compressed with sliders on the global page), and those are parts that appear darker if you apply local highlights slider to them. Highlights control work on the similar principle as highlights and shadows in PS and it's not intended to be used on very small areas like you did - it is (as I believe) based on a kind of wide area USM and you need larger area for it to work properly

                      My experience with it: same phenomenon in darker parts of moon, and independent of brush size / area of painted region.

                       

                       

                       

                      Vit Novak wrote:

                       

                      About "extreme" positions I mean global sliders.

                      As previously mentioned, the problem also occurs at global settings that are more tenuous.

                       

                       

                       

                      Vit Novak wrote:

                       

                      Highlights are not blown in raw file, but are highly compressed after whitebalancing and color space conversion and and further with your exposure slider setting (that's how PV2012 works - it rarely clips highlights, because a kind of soft clipping / compression is used instead of hard clipping in PV2003 at some places in the workflow)

                      For the record: there is zero clipping in PV2003/10 when all sliders are at zero (I mean factory default values) (using Camera Neutral v4 profile), but I'm not sure how relevant that is to the topic.

                       

                      Yes, the whites  are very compressed, thus the global -highlights and -whites. But even if you set all global basics at 0 the problem is still very-much evident.

                       

                       

                      So the questions still remain, in my mind anyway:

                      1. why is initial (painted in previous version) region so gray, compared to color obtained by repainting in current version.

                      2. why does even a very small amount of local -highlights have such an extreme effect, relative to global -highlights or (global or) local exposure...

                       

                       

                       

                      Rob

                      • 9. Re: PV2012 locals sometimes don't behave well...
                        Vit Novak Level 3

                        Aha, so by version you mean PV and not something elese, specific to LR. Still I'm not getting your 1st question

                         

                        About the second one - if I erase your mask and make a new one over the whole moon, I'm getting similar results if I set local highlights to -30 as if I increase global highlights from -70 to -100 (I'm rarely using something over -20 btw) - whole moon gets darker by similar amount. With your mask, only brightest parts of the moon are painted and affected with that slider. So I'm not getting what is wrong with behaviour of local settings

                        • 10. Re: PV2012 locals sometimes don't behave well...
                          areohbee Level 5

                          Previous text here was deleted...

                           

                          [edit]:

                          OK, after more experimental painting, I see that the grayness is really only in the parts that were nearest to over-exposure - which was the *only* part painted initially (I think you tried to tell me that - sorry it took so long for me to arrive there...). Still, I'm not sure why the color would be so different in the brightest parts like that, I mean *none* of the channels were blown, right? Oh well, at least I have a better handle on it than I did before.

                           

                          Question: it was near-to-impossible to get the paint to lay down the same way as originally painted - has there been some change to the paint-flow algorithm, or was it just me...?

                          [/edit]

                           

                          [edit #2]:

                          OK, so I was pretty much wrong about part 2 also, wasn't I? - local effect and strength is close to the global... Cr@p - I hate being wrong! . The bottom line is that the behavior of highlights slider when the tonal distribution is as it is in this photo is very "squirrelly" (for lack of a better word) - highlights adjustment is extremely sensitive and reaches way past the midline into the shadows - way more so than exposure does, which seems counter-intuitive, but is not entirely out of character for pv12, this is just a very extreme case...

                          [/edit #2]

                           

                          Anyway, here are some snapshots which illustrate the phenomenon

                          http://www.robcole.com/Rob/Personal/Pictures/LrForumSupport.cfm?embedSWF&openDir=Bad%20Moo n&openFile=20120505RC193027NB4328.O._v_Copy%201.jpg

                           

                          Note: you may have to context-click the link and choose "Open Link in New...".

                          Tip: use the big button, the full button, and the i button; click the next button repeatedly to compare.

                           

                          [edit #3]:

                          Original settings, with snail trails:

                          _bad_moon_settings_with_snail_trails.gif

                          note: Global basics did most of the heavy lifting, but there is also a subtle tone curve and some other subtle adjustments.

                           

                          As finished:

                          _bad_moon_settings_finished.gif

                          note: Global basics did most of the heavy lifting, but there is also a subtle tone curve and some other subtle adjustments.

                           

                          Note: Nikon D300, ISO2000.

                           

                          Explanation:

                          * Exposure was brought up to brighten the sky without relying soley on shadows slider.

                          * Contrast was brought down to decompress/debrighten moon and decompress/brighten dark tones: note: there are no midtones, per se, and contrast could be even more negative, but if too negative, there is a loss of richness that is hard to regain using +vib/sat/clarity...

                          * Highlights negative to de-brighten the moon, which was already over-bright, and to compensate for raised exposure.

                          * Shadows positive to assist exposure in brightening the sky.

                          * Whites negative to keep moon from being over-bright, -highlights alone does not suffice... (moon had "vividness" to spare, so to speak, and a little "dulling" via -whites was actually a plus in this case). Worth noting: -highlights darkens the sky too, a lot (even though not in the highlight tonal range as most people would think of it), -whites: not so much. It was primarily the "hyper-sensitive" (and non intuitive) behavior of the highlights slider, which made this photo such a challenge to edit in PV2012 (original edit was during Lr4 beta, before I had as much experience with pv12).

                          * Blacks must be negative to keep trees from being under-black and over-noisy - I wanted them to be silohuette black... Also, -blacks had the (in this case fortunate) side effect of bringing out (exagerating) the blue in the night sky.

                          * Clarity primarily to enhance moon detail (could have used local on the moon, e.g. radial gradient in Lr5). There was already sufficient tree/sky separation due to the -blacks +shadows combo, and one of the subtle locals was to decrease shadow strength around the tree/sky line, for a more natural look.

                          * Saturation because, well, I like color... (and strong negative contrast robs some color, no pun intended).

                           

                          Note: I experimented with "less extreme" settings, and there were other combos that were ok-ish, but without strong adjustments, as I was able to make them, this photo just couldn't make the cut, as I judged it, and would have been discarded.

                           

                          [/edit]

                           

                          PS - I am open to seeing an interpretation of this photo which employs "less extreme" (more moderate) settings, if anybody else is willing to give it a whorl...

                           

                          ~R.

                           

                          Message was edited (3 times) by: Rob Cole

                           

                          PPS - big thanks to Vit .

                          • 11. Re: PV2012 locals sometimes don't behave well...
                            Vit Novak Level 3

                            Yes, this photo is tough. Main problem is big difference in brightness of trees and sky vs moon, far exceeding dynamic range of output media

                             

                            As I said, highlights are very compressed during the ACR/LR workflow, so to get acceptable contrast there (and avoid moon looking bland and detailless), raising global exposure slider is no go. Unfortunately, lowering it  doesn't produce very good result either (although it should result in straightening upper part of effective total tone curve) and moon still looks bad, so I got better result with curves. Starting from all settings default, I set point curve, with control points

                             

                            (106,72)

                            (209,143)

                            (249,225)

                             

                            That improved appearance of the moon. Tried setting RGB or R, G, B individually that way - contrast is similar in both cases, but color is a bit different (obviously due to the fact how these curves are implemented - RGB is probably applied like tone curve of the profile). Second option looked better to me

                             

                            Since background is too dark, I added shadows +50. It's a matter of the taste how much to use - I generally don't like HDR look of photos, so this amount looked reasonable to me. Sky is less noisy than in your version and noise of the threes is acceptable to me. You can add more to be closer to the look of your photo

                             

                            WB 5000 K tint 0

                             

                            Added clarity +10, saturation +10

                             

                            Added some noise reduction

                             

                            Didn't use local adjustments (except global shadows, which is already a tool to increase local contrast, like highlights)

                             

                            Here is the result I'm relatively satisfied with

                             

                            Moon2b.jpg

                             

                            Noticed strange issue with WB here - raising it too high (10000K or so) results in green looking moon ... I suppose it has something to do with usage of Photo Pro as intermediate color space, like other similar issues with strange colors, but I'm not sure ...

                            • 12. Re: PV2012 locals sometimes don't behave well...
                              areohbee Level 5

                              Hey Vit,

                               

                              Initial reaction:

                              ------------------

                              Good: the moon looks GREAT!!!

                              Bad: I can't much see the trees against the sky on my monitor.

                              ------------------

                               

                              Stand by for more feedback (will have a play with it shortly, esp. shadows).

                               

                              Thanks again,

                              Rob

                              • 13. Re: PV2012 locals sometimes don't behave well...
                                Level 5

                                Rob Cole wrote:

                                 

                                …Bad: I can't much see the trees against the sky on my monitor…

                                 

                                Your monitor needs calibrating and profiling, Rob… or replacing. 

                                • 14. Re: PV2012 locals sometimes don't behave well...
                                  Level 5

                                  Incidentally, Bob, remember that this forum strips any embedded profiles in uploaded images, so if you just dragged it to  your desktop from the browser, you have an untagged file.

                                   

                                  Try assigning a different profile and see if that unblocks the shadows a bit for you.

                                   

                                  I also realize that few monitors can compete with the display of blacks I get from my ancient, but high-end CRTs. 

                                  • 15. Re: PV2012 locals sometimes don't behave well...
                                    RASouthworth Level 3

                                    I find that if I expand the picture on my NEC EA231WMi (LCD) the trees show up better.  Best results are from saving it off and opening in Photoshop, it really is a striking image.

                                     

                                    Richard Southworth

                                    • 16. Re: PV2012 locals sometimes don't behave well...
                                      RASouthworth Level 3

                                      Can't help myself, this is an image that begs to be dinked.  Left the moon alone, played with the background.  Not sying my version is better, just more to my taste.

                                       

                                      Richard Southworth

                                      image_jpegA.jpg

                                      • 17. Re: PV2012 locals sometimes don't behave well...
                                        areohbee Level 5

                                        Vit Novak wrote:

                                         

                                        (106,72)

                                        (209,143)

                                        (249,225)

                                         

                                        That improved appearance of the moon. Tried setting RGB or R, G, B individually that way - contrast is similar in both cases, but color is a bit different (obviously due to the fact how these curves are implemented - RGB is probably applied like tone curve of the profile). Second option looked better to me

                                         

                                        Indeed. Using the tone curve to drop and separate the highlights works better for this photo than the highlights slider. Far less +exposure is needed when -highlights is not used (and -whites not necessary...), and the moon highlights do not get jammed. I still don't understand why the result are so different using point curve vs. channel curves - I'll need to cogitate on that a while...

                                         

                                         

                                         

                                        Vit Novak wrote:

                                         

                                        Since background is too dark, I added shadows +50. It's a matter of the taste how much to use - I generally don't like HDR look of photos, so this amount looked reasonable to me. Sky is less noisy than in your version and noise of the threes is acceptable to me. You can add more to be closer to the look of your photo

                                        Here is a version that is similar to yours (based on same exact tone curve, but more intra-shadow contrast... (-blacks +shadows...):

                                         

                                        http://www.robcole.com/Rob/Personal/Pictures/LrForumSupport.cfm?embedSWF&openDir=Bad%20Moo n&openFile=20120505RC193027NB4328._v_Copy%201.jpg

                                         

                                         

                                         

                                        I kinda liked the candied over-the-top look of the previous version too, but I think overall I like the more natural version better.

                                         

                                        Thanks for all your help,

                                        Rob

                                        • 18. Re: PV2012 locals sometimes don't behave well...
                                          Level 5

                                          What I find amazing is the almost unbelievable latitude that this raw file provides in ACR 8.1RC.

                                           

                                          No way would I presume to do remotely as well as you, Rob, or Vit.  I also wasn't there and have no insight as to what you actually saw that night nor what your vision of the photograph is.  The following rendering is just one of many that are possible. All done in ACR 8.1RC, , no masking, no painting, no Lr and no Photoshop:

                                           

                                          20120505RC193027NB4328.O.jpg