Lightroom 4: The (Highlight) Wall

  • 1
  • Question
  • Updated 6 years ago
Some photos seem to have a wall up in the highlights.

Attempting to stretch highlights rightward e.g. via:

Exposure
Whites

results in tones ganging up against the wall - instead of stretching right-ward.

Here are some nice settings, but the image is not using the whole dynamic range:


Here is an attempt to expand it to use more of the upper end of the histogram:


The image was brightened, but did not expand into the right end of the histogram. Instead, it bunched up against a virtual "wall".

Even with extreme settings, the wall moves right-ward a little, but there is still a wall:


What's up with that???

Rob
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 372 Reply Likes

Posted 6 years ago

  • 1
Photo of Dorin Nicolaescu-Musteață

Dorin Nicolaescu-Musteață, Champion

  • 703 Posts
  • 38 Reply Likes
Can we have a file to play with?

So far, the only unusual thing to notice (from the histograms) is that your "highlights" are rather red.
Photo of Sebadja

Sebadja

  • 17 Posts
  • 0 Reply Likes
I assume, that there is something wrong with your tone curve. Look at the upper-right end of the curve. Is the last point really anchored to the top of the Curves box? I think, it is not. But if it is (against expectations), then try to push it to the left while making sure that you keep this point anchored to the top of the Curves box.
Photo of LRuserXY

LRuserXY

  • 426 Posts
  • 41 Reply Likes
Yes, tone curve could be the cause. I suspected in another post here that the tone curve comes last in the ACR processing pipeline (or at least *after* the basic tone controls), so cranking up the highlights/whites in basic will not compensate for lowering them in the tone curve.

Edit: Here is my other post (still unanswered): http://feedback.photoshop.com/photosh...

Edit 2: Here is my (extreme) example:



The cropped region already contains large areas of highlights in the 90%...98% range (all three channels) when using adobe defaults.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 372 Reply Likes
Tone/point curve is virginal.
Photo of LRuserXY

LRuserXY

  • 426 Posts
  • 41 Reply Likes
Strange... ok, must be some other effect, then. Is the "wall" still not moving when you set exposure to maximum?

And just to be sure: Is the tone curve neutral in the R/G/B channels, too? (Edit: Forget that question, if it was not, you would have noticed a color cast probably. And the name of the point curve preset would not read "linear").
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 371 Reply Likes
One can get everything mashed up into clipping if cranking exposure or whites all the way up.
Photo of LRuserXY

LRuserXY

  • 426 Posts
  • 41 Reply Likes
Ok, then my example was indeed pretty useless. So the question is why LR builds a wall (or at least displays a wall) for your picture or pictures of the same kind.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 371 Reply Likes
See post below - it is behaving as expected, I guess, given design and presence of tones not being shown(?)
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 372 Reply Likes
I'm not sure why Lightroom histogram does not show all the tones present in a photograph.

Looking at the histogram in NX2 reveals the answer:


ACDSee also shows high tones that Lightroom does not:


Conclusion:
---------------
There are tones in the upper-most region that are influential, but are not being shown by Lightroom. That, coupled with PV2012's highlight roll-off handling and such is the "culprit".

PS - In the past, I have noticed that such conditions ("over-extended tones") are also responsible for seeming underexposure when auto-toning in Lr3 - probably Lr4 too.

R
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 372 Reply Likes
Thanks LRuser24 - I guess MelissaRGB is the cause somehow (???), since switching to soft-proof mode shows the correct histogram - thanks for the tip.

-R
Photo of Dorin Nicolaescu-Musteață

Dorin Nicolaescu-Musteață, Champion

  • 703 Posts
  • 38 Reply Likes
What's the working space in NX and ACDSee?
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 372 Reply Likes
NX2 - AdobeRGB (I think), as I have it configured. Not sure 'bout ACDSee.
Photo of Dorin Nicolaescu-Musteață

Dorin Nicolaescu-Musteață, Champion

  • 703 Posts
  • 38 Reply Likes
Ok, then you cannot compare the histograms. Should be the same space.

Can you share the file? A reduced resolution dng should do.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 372 Reply Likes
I sent you a link via PM.
Photo of Dorin Nicolaescu-Musteață

Dorin Nicolaescu-Musteață, Champion

  • 703 Posts
  • 38 Reply Likes
Got the file. Thanks.
Hm.. very interesting.

The image in its initial state (first screenshot) has its red channel already blown in the AdobeRGB space. So, the basic controls seem to be programmed to, yes, use up dynamic range, but to a practical limit. And when you have highly saturated colors, that limit appears to be the AdobeRGB gamut.

And what would be the point to push further, we don't have a lot of output devices with a gamut wider than that, do we?

If you still want to use up all your "theorectical" dynamic range - there's the point curve.

EDIT: what confused you (and us) in this particular case is that "theoretical" color space used for the histogram - ProPhoto.
Photo of Andrew Rodney

Andrew Rodney

  • 476 Posts
  • 76 Reply Likes
>what confused you (and us) in this particular case is that "theoretical" color space used for the histogram - ProPhoto.

Exactly, blue and green fall outside human vision (CIE XYZ).
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 372 Reply Likes
It does seem MelissaRGB is closer to ProPhotoRGB than AdobeRGB or sRGB. It's a mystery to me why ProPhoto & Melissa suppress colors that are shown by Adobe/s. If anything, it seems the opposite would be true. I'm clearly outside my area of expertise now...
Photo of Lee Jay

Lee Jay

  • 990 Posts
  • 131 Reply Likes
Melissa is ProPhoto primaries with an SRGB tone curve.
Photo of Eric Chan

Eric Chan, Camera Raw Engineer

  • 616 Posts
  • 119 Reply Likes
Correct, the Develop histogram's color primaries are the same as ProPhoto.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 372 Reply Likes
Well, something seems non-optimal here. I mean, tones are influencing the editor, which I can't see in the histogram. It seems that it would be best to display all tones that are affecting slider behavior.
Photo of Dorin Nicolaescu-Musteață

Dorin Nicolaescu-Musteață, Champion

  • 703 Posts
  • 38 Reply Likes
You see it i the histogram. What's confusing, is that you (an me too) are expecting the algorithms to use up all available dynamic range in the working space. But they don't have to, even if they can!

Instead, the math is optimized to not totally blow your sunset red out of a reasonable AdobeRGB gamut. At least that's how I understand it...
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 372 Reply Likes
I only see it in the histogram if I switch to soft-proof mode, or load it into NX2 or ACDSee. I don't otherwise see it in Develop Module histogram (see posts with histogram examples above).
Photo of LRuserXY

LRuserXY

  • 426 Posts
  • 41 Reply Likes
...because LR "refuses" to generate these tones in order to keep them in a reasonable gamut, as Dorin pointed out - I think that could be the cause, too.

Example (try it in Photoshop): A maximum saturated and maximum bright Red in AdobeRGB (255/0/0) when converted to ProPhotoRGB becomes 216/85/32 in ProPhotoRGB. If the red channel in ProphotoRGB is set higher than 216, the color goes out of the AdobeRGB gamut (even if the other channels are increased, too!!). 216 would be the "wall" in this example. So the theory is that LR tries to "contain" the colors inside AdobeRGB (or some other predefined "border color space"), which of course also depends on their hue, saturation and luminance (try some other values in Photoshop).

P.S. The example in ProPhotoRGB is not exactly what happens in Lightroom because the histogram is in MelissaRGB, not ProPhotoRGB (different tone curve). However, in principle it is the same.