Lightroom: A new record for most over-exposed auto-toning.

  • 2
  • Problem
  • Updated 5 years ago
  • Solved

win7/64, Lr4.4RC
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes

Posted 5 years ago

  • 2
Photo of jdv

jdv, Champion

  • 728 Posts
  • 56 Reply Likes
Since you don't supply an example photo, this is marginally useful at best.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Sorry for being so marginally useful - but, right back at ya.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Things to note:

* Exposure is set beyond the maximum!
.
Photo of Chris Cox

Chris Cox

  • 20280 Posts
  • 813 Reply Likes
Yes, but what image got it there? Or is this just fun with Photoshop?
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Lightroom - I think the engineers/programmers should be aware that the auto-toner has the capacity to exceed the maximum that has been set for the exposure setting - please pass along this information to them, if they are not reading this forum themselves. - thanks. I don't have permission to share photo.
Photo of Chris Cox

Chris Cox

  • 20280 Posts
  • 813 Reply Likes
Without an image that causes the over-range case, they'll just kick it back as "cannot reproduce". Even if you have to obfuscate the image or create a new one - they need an example to figure out how they got into this state.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
In case the purpose for this thread is still not clear:

1. Auto-toner still massively over-exposes many photos, and very-much under-exposes some too. If you can't find examples to work with, one guy posted several hundred, in response to Eric Chan's "seems like auto-toned exposure is working pretty good to me" reply.

2. The exposure result is not being tested against the maximum constraint - I bet one of the Lr programmers can find the reason by inspecting the code... (hint: if not readily obvious: add some debugging logic).

Rob
Photo of Chris Cox

Chris Cox

  • 20280 Posts
  • 813 Reply Likes
I'd guess that exceeding the UI constant is by design. But going that far probably isn't. Again, they will almost certainly need an image that reproduces the behavior.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
I'd guess having a final setting's value exceed the maximum is a bug. Hopefully, knowing that such bug exists is more useful than not knowing. I have now provided such information to you, to do with as you see fit. As I said, I don't have permission to post the photo, but it was a D300 NEF file, with a strong right-ward extending red channel, under-exposed on purpose in the interest of not clipping the red channel, but under-exposed too much since such clipping was being judged by the in-camera jpeg - the raw wasn't so near to clipping as the in-camera jpeg was indicating (I've learned a few things since that photo was taken ;-}). In addition to the bright red light, there was a fair amount of dark shadow in the photo as well. Hope that helps.
Photo of Chris Cox

Chris Cox

  • 20280 Posts
  • 813 Reply Likes
Could you create a similar photo that gives similar results?
Again, just that knowledge may not be enough to find the fault.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
In previous post, I mentioned 2 problems.

The first is, by far, the most poignant. Maybe if they go deep into solving #1 they'll find the solution to #2 in the process. If not, just having #1 solved would be a dream come true for many auto-tone users, or auto-tone user wannabes.

Rob
Photo of Chris Cox

Chris Cox

  • 20280 Posts
  • 813 Reply Likes
Again, I suspect that it is intended to go beyond the UI limits.
The real problem is how it got that FAR beyond.
And to figure that out, they will need an image that reproduces the problem.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Again, the real problem is the over-exposure in so many cases - not the rare case when it exceeds 5.0. Nevertheless, I will see about sending another sample (I, along with many others, have already sent quite a few samples).
Photo of Chris Cox

Chris Cox

  • 20280 Posts
  • 813 Reply Likes
Thank you.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
cc of recent email to Eric Chan:

Hi Eric,

This photo (attached) is especially prone to over-exposure by auto-toning.

Also note:

• Auto-toned exposure exceeds 5.0 (is +10.00) for all camera profiles I tried except *plain* Adobe Standard (e.g. it happened with my customized version of Adobe Standard and some v4 camera emulation profiles).

Hope this helps, and please let me know if there is anything else I can do.
Rob
Photo of Allan Olesen

Allan Olesen

  • 64 Posts
  • 6 Reply Likes
Letting the exposure go past the slider maximum seems like an active decision to me, not a bug.

My reasoning:
They have not just increased exposure past the slider maximum. That could easily happen.
But they have ALSO updated the UI for the slider to reflect that value. That does not happen by accident.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Note: if one attempts to adjust exposure down a little from the 10 at which it is set by the auto-toner, it goes immediately to 4.9X, without apparent exposure going down more than a little.

What that says to me is that the auto-toner had set the value only a little past 5 internally, and the 10 that got displayed was just an inconsequential UI glitch (let's call it a bug).

I can't stress this enough:

The real problem here is *not* that the value that got displayed was above the expected maximum, the problem is that the auto-toner radically overexposes some (many) shots.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Note: it's ~impossible to drive this photo into clipping. At an exposure of about 1.5, it starts "climbing a virtual wall" about a "half inch" before the right edge of the histogram. Then, if pushed via either exposure or whites or both, just keeps climbing the wall, pushing it rightward some but never clips, not even close, period, even at exposure=5 and whites=100:

After reset, auto-tone, and exposure reduced to 1.5:


With exposure doubled to about 3:


And with exposure max'd at 5, and whites all the way to 100:


I suspect the auto-toning algorithm just keeps uping exposure in the hopes of seeing tones get near to clipping, without taking into account some of the characteristics of PV2012 (one of which I call "the wall", short for "premature highlight ganger-upper wall") that prevent it from happening...

Rob
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Here's a lossy dng if anybody wants to play:

http://www.robcole.com/LrForumSupport...

Note: you have to right/ctrl-click for context menu, then choose "Save Link As..." (or whatever is closest in your OS).

Reminder: choose a v4 camera emulation profile if you want to see auto-toned exposure set past +5.0 (erroneously showing +10.0, although it's not really that high).
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Status update: I never heard back from Eric, whether he got the file I sent, or downloaded it...
Photo of Jeffrey Tranberry

Jeffrey Tranberry, Sr. Product Manager, Digital Imaging

  • 14049 Posts
  • 1735 Reply Likes
I pinged Eric.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Thanks Jeffrey. - standing by...
~R.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Any news?
R
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
?
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 379 Reply Likes
Much better in Lr5.2 - instead of 10.0 (or ~5.0 using Adobe Standard), auto-toner is setting exposure to 2 or 2.5 (depending on camera profile) - a reasonable value.