ETTR, Blacks & Exposure settings vs. Point Curve controls

[Peter_DL]

 
Initially I wanted to add this to another current discussion, then I was not sure if and how to post this at all, but then it is quite ACR specific topic and could be worthwhile for an own discussion:

In the context of ETTR, or maybe in general, there are some comparisons being of interest with regard to ACR’s choice of tools. Although there are a couple of sub-cases depending on image, and interactions between these cases, these basic questions are as follows:

1.)  minus-Exposure setting  vs. white point setting via the Point Curve.
2.)  same question but in the plus-Exposure direction. Let’s leave it out here.
3.)  plus-Blacks setting  vs. black point setting via the Point Curve.

Example:

The first two shots / screenshots as given below are camera exposed +/- one stop apart. The delta as shot is 2 EV. In ACR, both images were normalized with Exposure & Blacks sliders to RGB 190 for the second brightest gray and RGB 64 for the second darkest gray (referring to ProPhoto RGB readings, with a precision of +/-2 level per channel). All other tonal controls were zeroed. Baseline matrix used.

The surprising thing is that the second shot, the ETTRed image, also needs higher setting with the Blacks slider, even and in particular after an initial normalization with the Exposure slider. Actually, I’m not the first one who found that ETTRed images call for a quite excessive setting with the Blacks slider and some people have complained about an excess of saturation received. Then, it is sometimes recommended to resort to black point setting via the Point Curve. Hmm.

The more surprising thing (for me) is that this difference between the Blacks setting between both images (i.e.  Blacks 6 vs. 20) is remaining without consequence on color saturation. Whereas with one single image, it can be easily shown that a move from 6 to 20 makes a pronounced effect, increasing color saturation in the shadows with decreasing significance to the lights – as expected from plain RGB math. For the purpose of illustration, image # 1 (the darker one) got increased from Blacks 6 to 20. Results are given with image # 3.

Preliminary conclusion (after some further tests):  black point setting via the Point Curve tab sounds cool, but it seems to add nothing to the party.

Thoughts, insights ?

Peter

--

[stamper]

I don't think you should worry too much about what goes on beneath the hood unless of course you are one of the ones  - a few on here - who seem to be obsessed with the theory rather than the practical side of things. If you like the final result after tweaking the various sliders and it doesn't produce extra noise then be happy. As to saturation increases that happens in PS when you alter the curves but can be sorted by changing the blend mode to luminosity. You can't do this in ACR but the good news is you can reduce the saturation in different colours. For every gain there is also a downside but in ACR it can usually be rectified easily.

[ErikKaffehr]

Hi,

I'd say that it matters little how normalization is done for the blacks. Regarding highlights I strongly feel that expsure would be used.

With the old processing pipeline I sometimes got posterisation in the dark shadow area with the old processing pipeline in LR, using the new pipeline I'd think there may be less problems. Using the point curve to supress blacks instead of the black slider helped on the issue. But this was an issue I had on a single exposure of 20000 (or so).

Best regards
Erik

 
Initially I wanted to add this to another current discussion, then I was not sure if and how to post this at all, but then it is quite ACR specific topic and could be worthwhile for an own discussion:

In the context of ETTR, or maybe in general, there are some comparisons being of interest with regard to ACR’s choice of tools. Although there are a couple of sub-cases depending on image, and interactions between these cases, these basic questions are as follows:

1.)  minus-Exposure setting  vs. white point setting via the Point Curve.
2.)  same question but in the plus-Exposure direction. Let’s leave it out here.
3.)  plus-Blacks setting  vs. black point setting via the Point Curve.

Example:

The first two shots / screenshots as given below are camera exposed +/- one stop apart. The delta as shot is 2 EV. In ACR, both images were normalized with Exposure & Blacks sliders to RGB 190 for the second brightest gray and RGB 64 for the second darkest gray (referring to ProPhoto RGB readings, with a precision of +/-2 level per channel). All other tonal controls were zeroed. Baseline matrix used.

The surprising thing is that the second shot, the ETTRed image, also needs higher setting with the Blacks slider, even and in particular after an initial normalization with the Exposure slider. Actually, I’m not the first one who found that ETTRed images call for a quite excessive setting with the Blacks slider and some people have complained about an excess of saturation received. Then, it is sometimes recommended to resort to black point setting via the Point Curve. Hmm.

The more surprising thing (for me) is that this difference between the Blacks setting between both images (i.e.  Blacks 6 vs. 20) is remaining without consequence on color saturation. Whereas with one single image, it can be easily shown that a move from 6 to 20 makes a pronounced effect, increasing color saturation in the shadows with decreasing significance to the lights – as expected from plain RGB math. For the purpose of illustration, image # 1 (the darker one) got increased from Blacks 6 to 20. Results are given with image # 3.

Preliminary conclusion (after some further tests):  black point setting via the Point Curve tab sounds cool, but it seems to add nothing to the party.

Thoughts, insights ?

Peter

--

[digitaldog]

3.)  plus-Blacks setting  vs. black point setting via the Point Curve.

ETTR not withstanding, I prefer using the black point setting via curves (the Shadows slider). I have no idea what the differences are, this is strictly subjective. George Jardine, formally of Adobe LR Team showed me this ‘trick’ and I’ve been using it since then. I just like the way the image looks.

[madmanchan]

Hi Peter, the reason for what you're seeing is that currently ACR's Blacks control does not account for Exposure normalization.  So the exact impact of a given Blacks slider value (e.g., 5) is highly dependent on how you exposed the image originally, regardless of whatever software Exposure compensation you apply in post-processing. 

For example, shoot a scene in raw mode at a given exposure.  Shoot the exact same scene again, but 2 stops darker (e.g., 1/4 the exposure time).  Open both images in ACR with default settings, including Blacks 5.  Set the Exposure of the darker one to +2 (to compensate for the 2-stop darker exposure at capture time).  They won't look the same.  The 2nd image will have much darker shadows (and by side effect, increased saturation and contrast).  This is because Blacks=5 has a much stronger effect on the 2nd darker-exposed image, compared to the 1st much-brighter image.  The opposite thing happens if you run the experiment in the other direction (i.e., 2nd image is 2 stops brighter -- you'll find that Blacks=5 has much less effect). 

If you set Blacks to 0 for both images, then you'll find that they look the same (after adjusting Exposure in the 2nd image to +2, of course).

[Peter_DL]

... Regarding highlights I strongly feel that expsure would be used.

Yes,
if we isolate such comparison which corresponds to 1.)
minus-Exposure setting  vs. white point setting via the Point Curve,
there are three aspects which come in my mind:

a.)  The Exposure slider, unlike white point setting via the Point Curve tab, won’t cut spectacular highlights which are truly clipped down from RGB 255 to a muddy gray. So the Exposure slider will be preferred.

b.)  There will be "irregularities" / differences in the top highlights due the smoothening of the Exposure function as well as due to ACR’s recovery strategy which was shown to extend to non-Raw-clipped areas.

So far this was discussed in the earlier thread quoted above.

c.)  Leaving out such highlight scenarios, following test shows another difference (see screenshots below):   Imange 1 # 01, the darker shot (- 1 EV), was left at an unaltered "linear" stage. Imange 1 # 02, the bigther shot (+ 1 EV), was Exposure-downadjusted in ACR to the same value with the second brightest gray patch. Whereas with Imange 1 # 03 this Exposure adjustment was replaced by a corresponding white point setting via Point Curve.

Imange 1 # 01 and # 02 appear to be the same, whereas Imange 1 # 03 looks noticeable brighter.
Apparent conclusion: white point setting via Point Curve leaves the rest of the image brighter than a corresponding minus-Exposure setting.

Peter

P.S.:  Just saw the additional replies and will come back tomorrow.
Many thanks.

--

[stamper]

Quote Eric

Set the Exposure of the darker one to +2 (to compensate for the 2-stop darker exposure at capture time).

Unquote

Does this mean that the exposure setting in ACR has a direct correlation with the + EV and - EV on your camera? Recently when I had been using + EV in camera and adjusting the exposure down in ACR I noticed that a +1 EV meant that roughly a - 1 exposure was needed. Could I then set it exactly -1 instead of a subjective guess at it?

[pegelli]

Stamper, that's what I have always assumed. I did a quick test a while back in lightroom. First I zeroed everything with a straight linear curve. Subjectively my +1 EV in camera/ -1EV in lightroom looked exactly the same as a fully uncorrected (in camera/LR).
Hovering over some areas the % R, G and B as well as the histogram was nearly identical.

Must say the scene didn't have strong highlights nor strong/dark shadows, so the comparison was more in the middle of the histogram (10 - 90%)

[stamper]

I too zero my settings before processing. I wouldn't have guessed this. I have Jeff's book on the subject and I don't remember it being there. It is a while since I read it so my ageing memory may be at fault. It is interesting to know this snippet of information.

[Peter_DL]

....

Please bear with me regarding this supplement to post # 5.
It is the inverse test here, which corresponds to the initial point 2.)
plus-Exposure setting  vs. white point setting via the Point Curve.

Imange 2 # 01 as given below, which is based on the darker shot (- 1 EV), was Exposure-upadjusted in ACR to RGB 190 with the second brightest gray patch. With imange 2 # 02, this Exposure adjustment was replaced by a corresponding white point setting via Point Curve. Image 2 # 03, the brighter shot (+ 1 EV), was essentially left at an unaltered "linear" stage, with just a minimal Exposure adjustment to meet RGB 190.

The screenshots 2 # 01 and # 03 look like the same, whereas Imange 1 # 02 (the one with point curve setting) looks noticeable darker. Apparent conclusion: white point setting via Point Curve leaves the rest of the image darker than a corresponding plus-Exposure setting.

More general:  White point setting via Point Curve has less (side-) effect on the midtones and shadows than +/-Exposure setting in ACR.

Unexpectedly, since ACR’s Exposure slider was shown to behave quite linear. Guillermo had an excellent chart for illustration. Whereas white point setting via Point Curve would also suggest a linear relationship between input and output.
However, the empirical finding here is in line with another earlier discussion and what was reported by other user. Unfortunately the illustration on Michael Frye’s website does not seem to be online anymore.


Regards, Peter

--

[madmanchan]

Does this mean that the exposure setting in ACR has a direct correlation with the + EV and - EV on your camera? Recently when I had been using + EV in camera and adjusting the exposure down in ACR I noticed that a +1 EV meant that roughly a - 1 exposure was needed. Could I then set it exactly -1 instead of a subjective guess at it?

Yes, the Exposure scale in ACR is measured in stops, with the same photographic and physical meaning as in the camera.

However, there are some practical differences, of course.  For example, if you reduce the capture exposure in the field (e.g., -1 stop by using 1/60 sec instead of 1/30 sec), then you can record new information in the highlights that previously would have been clipped and forever lost. In other words, the top stop of your shorter exposure (1/60 sec) may have some new highlight detail in it, which the longer exposure (1/30 sec) would not have had.  In contrast, if you bring the longer exposure into ACR and set Exposure slider to -1, the overall image brightness will match that of the 1/30 exposure, but you won't be able to recover lost highlight details in that top stop.

[madmanchan]

Hi Peter,

In ACR, the point curve is not applied in a linear space.  So it will have a fundamentally different effect on the distribution of tones in the image, compared to an Exposure control (which by design is intended to be applied in a linear space).

[Peter_DL]

Hi Peter,

In ACR, the point curve is not applied in a linear space.  So it will have a fundamentally different effect on the distribution of tones in the image, compared to an Exposure control (which by design is intended to be applied in a linear space).

Hi Eric,

Many thanks for confirmation / explanation !

In summary of 1.) and 2.) I can only imagine resorting to Point Curve White Point setting (with an image without spectacular, blown highlights), when - and IF – it is rendering an advantage in the top highlights. Otherwise, the somewhat different tonal response of the tools (as shown here) does not seem to be worthwhile to address to the point curve tab at this stage. There are other tools also in the Basic tab to adjust midtone brightness and shadows.
In other words, better stay with the Basic tab Exposure setting unless top highlight rendition would possibly tell the opposite.

Makes sense ?

Regards, Peter

--

[madmanchan]

Hi Peter,

In terms of highlight clipping (what's clipped, what's not), you're best off working with Exposure (or Recovery) in Basic.  Tone curve controls (Brightness & Contrast in Basic,
Parametric & Point in Tone Curve) don't influence highlight recovery at all, so once the data is gone, it's gone ... for example, lowering the white point with the point curve will simply create gray blobs.  It won't bring back any details in those white areas.

[Peter_DL]

Hi Peter, the reason for what you're seeing is that currently ACR's Blacks control does not account for Exposure normalization.  So the exact impact of a given Blacks slider value (e.g., 5) is highly dependent on how you exposed the image originally, regardless of whatever software Exposure compensation you apply in post-processing.  

For example, shoot a scene in raw mode at a given exposure.  Shoot the exact same scene again, but 2 stops darker (e.g., 1/4 the exposure time).  Open both images in ACR with default settings, including Blacks 5.  Set the Exposure of the darker one to +2 (to compensate for the 2-stop darker exposure at capture time).  They won't look the same.  The 2nd image will have much darker shadows (and by side effect, increased saturation and contrast).  This is because Blacks=5 has a much stronger effect on the 2nd darker-exposed image, compared to the 1st much-brighter image.  The opposite thing happens if you run the experiment in the other direction (i.e., 2nd image is 2 stops brighter -- you'll find that Blacks=5 has much less effect).  

If you set Blacks to 0 for both images, then you'll find that they look the same (after adjusting Exposure in the 2nd image to +2, of course).

Hi Eric, I think I got it now.
Many thanks for explanation.

So in conclusion, as an ETTR shooter we should be basically not afraid when an unusual high Blacks setting is needed in post-processing in ACR. It won’t eat the image, because the Blacks slider is then acting less "aggressive" than with a "normally", darker exposed shot. The Blacks slider is changing its effectiveness between differently camera-exposed shots, so that the setting values cannot be cross-compared. This aspect occasionally came up in the past, e.g. here and there.

Some clarifications reached above on the initial points 1.) and 2.) about "ACR exposure / wp setting"
also allow for a more careful investigation of point 3.) plus-Blacks setting  vs. black point setting via the Point Curve.

Image / screenshot  3 # 01, the darker capture (- 1 EV), was basically left again at an unaltered "linear" stage, but with an arbitrary Blacks 10 setting. Image 3 # 02, the brighter shot (+ 1 EV), was initially Exposure normalizated, then a Blacks setting of 40 was needed to match the overall brightness and to come to a numerical match with the second darkest gray.  Both images then look reasonably the same, no spoiled color saturation from this strong Blacks move.  With Image 3 # 03, this Blacks 40 setting was replaced by a corresponding black point setting via Point Curve, again numerically matched to the second darkest gray. This one seems to look somewhat brighter, particularly in the deeper shadows.

This result may deviate in parts from the initial example with post # 1. But then, we have been approaching more systematically since then (since post # 5). Also, it might fit better now to the insight that "in ACR, the point curve is not applied in a linear space".

Black point setting via Point Curve may possibly indicate an initial advantage regarding tonal distribution in the deep shadows, however, it may go beyond of the testing scope here if this couldn’t be reached likewise by combining Blacks with the other tonal adjustments. Same regarding black point setting via curves (i.e. the shadows slider of the parametric curve) as Andrew suggested. This aspect finally remains pretty much ETTR related, when an unusual high Blacks setting is needed in post-processing in ACR/LR, to try to replace it by Point Curve BP or Parametric Curve Shadows setting.

Regards, Peter

--

[Peter_DL]

Hi Peter,

In terms of highlight clipping (what's clipped, what's not), you're best off working with Exposure (or Recovery) in Basic.  Tone curve controls (Brightness & Contrast in Basic,
Parametric & Point in Tone Curve) don't influence highlight recovery at all, so once the data is gone, it's gone ... for example, lowering the white point with the point curve will simply create gray blobs.  It won't bring back any details in those white areas.

Hi Eric,

Coming back to the subject about top highlights (and while leaving the ColorChecker exercises as done above), you are of course right that minus-ACR-Exposure avoids to turn specular reflection spots from white to muddy gray as with Point Curve WP setting. However, there is another aspect I’d like to illustrate:

With the following image (which we had already seen earlier) and the crop given below there is a transition in from the non-Raw-clipped left part of the sky to Raw-channel-clipping at the right side. The initial "linear" rendition in ACR ranges from a red hue left to yellow on the right, due to the cut-off with the R channel and R=G>B.

If for instance we want to make the sky a bit darker, minus-Exposure does not necessarily deliver the more pleasing rendition.  Probably, it is due to the recovery mechanics jumping in and trying to recreate a red hue, however, in a less saturated fashion towards R=G=B which apparently also expands to the ok areas (non-Raw-clipped). The result may look a bit "boring", depending on taste. Whereas with Point Curve WP setting the initial color range of vivid red to yellow is maintained as it is, just brought to a lower tone level as intended.

Regards, Peter

--
P.S.: attached screenshots will probably need to be viewed and compared in a color-managed environment such as PS.