Ad
Ad
Ad
Pages: « 1 [2] 3 4 »   Bottom of Page
Print
Author Topic: Exposing To The Right  (Read 18791 times)
sandymc
Sr. Member
****
Offline Offline

Posts: 270


« Reply #20 on: August 13, 2011, 03:58:50 PM »
ReplyReply

I'm still looking for a demonstration of that colour shift.

Guillermo,

It's just the "usual" Adobe profile-with-hue-twists problem.

Regards,

Sandy
Logged
ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #21 on: August 13, 2011, 07:03:29 PM »
ReplyReply

Some relevant info in a related thread:

http://www.luminous-landscape.com/forum/index.php?topic=56906.msg460680#msg460680

I see two issues:
(1) Does ETTR make any difference in the quality of the raw data; 
(2) if one practices ETTR (or ETTL on cameras such as the D7000, K5, or any CCD sensor'd camera), does your raw converter of choice handle the data properly.

The answer to (1) depends on the camera, as discussed in detail in the post linked above.  As for (2), I think many raw converters don't do a good job of exposure correction; Adobe has the issues with hue twist mentioned by Sandy, and Guillermo has suggested in a parallel thread that DPP doesn't have a true exposure control.  Converters that definitely treat exposure compensation correctly are RPP and RawTherapee.
Logged

emil
bjanes
Sr. Member
****
Online Online

Posts: 2825



« Reply #22 on: August 13, 2011, 08:17:23 PM »
ReplyReply

Some relevant info in a related thread:

http://www.luminous-landscape.com/forum/index.php?topic=56906.msg460680#msg460680

I see two issues:
(1) Does ETTR make any difference in the quality of the raw data; 
(2) if one practices ETTR (or ETTL on cameras such as the D7000, K5, or any CCD sensor'd camera), does your raw converter of choice handle the data properly.

The answer to (1) depends on the camera, as discussed in detail in the post linked above.  As for (2), I think many raw converters don't do a good job of exposure correction; Adobe has the issues with hue twist mentioned by Sandy, and Guillermo has suggested in a parallel thread that DPP doesn't have a true exposure control.  Converters that definitely treat exposure compensation correctly are RPP and RawTherapee.

Hue twists are frequently attributed to ACR, but Eric Chan has stated in another thread that if one uses ETTR and normalization of the image is necessary, one should use the exposure slider first to get the tonal values in the proper range before the 3D color table is applied.

Regards,

Bill
Logged
schitti
Newbie
*
Offline Offline

Posts: 5


« Reply #23 on: August 15, 2011, 04:20:56 AM »
ReplyReply

Digitizing week signals is a well known problem in scientific instrumentation including increasing the signal to noise ratio by averaging of repeated signals.

I have worked all me live in the field of Nuclear Magnetic Resonance and optimizing gain for the actual digitizer was automated thirty years ago. Increase the gain as much as possible but do not saturate ( clip ) the digitizer ( and the complete receiving system ). Exactly the same is true for digital cameras.

 Yes, we had laboratory computers to solve this task at that time and I fully agree this feature could be well integrated in modern Cameras.

Werner Schittenhelm
Logged
Wayne Fox
Sr. Member
****
Offline Offline

Posts: 2886



WWW
« Reply #24 on: August 16, 2011, 01:09:01 AM »
ReplyReply

Hue twists are frequently attributed to ACR, but Eric Chan has stated in another thread that if one uses ETTR and normalization of the image is necessary, one should use the exposure slider first to get the tonal values in the proper range before the 3D color table is applied.

Regards,

Bill
I know this is the "proper" way to do it, but since the exposure slider isn't a true full linear control, I've gotten better results by setting my black point with the black slider first, then setting my white/highlight point with the exposure slider, and finally tweaking the image brightness with the brightness slider.  If I go in that order I almost always get a very pleasing starting point that requires very little tweaking with the recovery and fill light sliders. Sometimes when I set my exposure slider first I have trouble getting my shadows to look right, and my highlights don't look as good after setting the black slider.

Am I crazy?
Logged

sandymc
Sr. Member
****
Offline Offline

Posts: 270


« Reply #25 on: August 16, 2011, 06:02:25 AM »
ReplyReply

Hue twists are frequently attributed to ACR, but Eric Chan has stated in another thread that if one uses ETTR and normalization of the image is necessary, one should use the exposure slider first to get the tonal values in the proper range before the 3D color table is applied.

I don't often disagree with Eric, but that's simply not the case. Adobe color profiles have two color tables, not one - a "LookTable" and a "HueSatDelta" table. One is applied after exposure correction, and the other before. Eric's advice is good in the sense that the process he recommends minimizes the amount of hue change that will occur, and if all the hue twists are in the LookTable there will be no hue changes. But if the profile has a HueSatDelta with hue twists in it, you will still get hue changes.

Sandy
« Last Edit: August 17, 2011, 01:11:55 AM by sandymc » Logged
RFPhotography
Guest
« Reply #26 on: August 16, 2011, 08:03:27 AM »
ReplyReply


2) The histogram is conservative, as it is not calculated from raw data. So it's possible to push exposure a bit longer, but we don't actually now how far.

If you do testing with your camera, you can get a very good sense of how far you can push the exposure without clipping any channels.

Logged
ErikKaffehr
Sr. Member
****
Online Online

Posts: 7672


WWW
« Reply #27 on: August 16, 2011, 02:14:58 PM »
ReplyReply

Crazy? Not..

But... adjusting the image is an iterative process. The important thing is that "overexposure/ETTR" should be compensated with the exposure slider. Highlight recovery will induce artifacts.

For me, the exposure slide first approach is working well. As a side note, when doing panos I normally "maximize the image" essentially setting black level to zero and even add some fill light. "Normalization" I often do on the merged pano.

As pointed out many times, the order the adjustments are done is not relevant for processing, but has a significant effect on the way you work with the images.

Best regards
Erik

I know this is the "proper" way to do it, but since the exposure slider isn't a true full linear control, I've gotten better results by setting my black point with the black slider first, then setting my white/highlight point with the exposure slider, and finally tweaking the image brightness with the brightness slider.  If I go in that order I almost always get a very pleasing starting point that requires very little tweaking with the recovery and fill light sliders. Sometimes when I set my exposure slider first I have trouble getting my shadows to look right, and my highlights don't look as good after setting the black slider.

Am I crazy?
Logged

madmanchan
Sr. Member
****
Offline Offline

Posts: 2110


« Reply #28 on: August 16, 2011, 09:03:02 PM »
ReplyReply

No, the Exposure stage in ACR/LR happens before the 3D LookTable, not after.  Same with Blacks.  So if you use these to adjust the image (to compensate for ETTR) you should get the same result.  For example, compare two cases.  For Case A, I take an image at a given exposure. For Case B, I take the same image, but 1 stop underexposed (e.g., halve the shutter time), then set Exposure to +1 in ACR to make up for it.  Aside from shadow clipping due to the default Blacks = 5, the tones and colors in the results should be the same in both cases, even if using a lightness-dependent profile, e.g., Adobe Standard or one of the Camera Matching profiles.

I am also puzzled by the discussions of the non-linearity of ACR's Exposure.  ACR's Exposure in the positive direction is simply a straight multiply with a hard clip, just like digital camera exposure.  In the minus direction, the only difference is that ACR tries to keep clipped whites white (i.e., we did not feel it was photographically useful to let speculars turn into gray blobs, though we still let users accomplish that with the point curve).  Remember, when reducing (software) Exposure, there's nothing "above" the sensor saturation point (very unlike at capture time, where reducing the capture exposure can indeed record additional information), so there is a question of how to treat the whites.
Logged

sandymc
Sr. Member
****
Offline Offline

Posts: 270


« Reply #29 on: August 17, 2011, 01:22:47 AM »
ReplyReply

No, the Exposure stage in ACR/LR happens before the 3D LookTable, not after.  Same with Blacks.  So if you use these to adjust the image (to compensate for ETTR) you should get the same result.  For example, compare two cases.  For Case A, I take an image at a given exposure. For Case B, I take the same image, but 1 stop underexposed (e.g., halve the shutter time), then set Exposure to +1 in ACR to make up for it.  Aside from shadow clipping due to the default Blacks = 5, the tones and colors in the results should be the same in both cases, even if using a lightness-dependent profile, e.g., Adobe Standard or one of the Camera Matching profiles.

Yes, but there's still the other table (technically, the interpolated combination of ProfileHueSatMapData1 and ProfileHueSatMapData2). That's done before exposure compensation. Or otherwise someone should rewrite the profile spec real quick. And the way ACR and LR work Grin

Or are you saying that for all current Adobe profiles, hue twists are only ever in the LookTable, not the HueSatMap tables?

Regards,

Sandy
Logged
madmanchan
Sr. Member
****
Offline Offline

Posts: 2110


« Reply #30 on: August 17, 2011, 08:44:06 AM »
ReplyReply

Hi Sandy,

You are correct in that the HueSatMap1 and HueSatMap2 tables precede the Exposure & Blacks controls.  However, all Adobe-provided profiles (e.g., Adobe Standard, Camera Standard, Camera Faithful, etc.) only ever put 3D tables (i.e., lightness-dependent) into the LookTable tag. 

We actually discourage (but do not prevent) the use of 3D tables in the HueSatMap* tags for the reasons you mention.  The preferred use of the HueSatMap* tags is a 2.5D table, indexed by hue & sat only, to perform non-linear color corrections.  This can be important with cameras with suboptimal spectral sensitivities or when photographing under difficult illuminants.  Lightness-dependent corrections are not necessary because the input data is in a linear, scene-referred space.  The Chart Wizard feature of the DNG Profile Editor writes into the HueSatMap* tags and writes only 2.5D tables.

Eric
Logged

digitaldog
Sr. Member
****
Offline Offline

Posts: 9191



WWW
« Reply #31 on: August 17, 2011, 09:29:24 AM »
ReplyReply

We actually discourage (but do not prevent) the use of 3D tables in the HueSatMap* tags for the reasons you mention.  The preferred use of the HueSatMap* tags is a 2.5D table, indexed by hue & sat only, to perform non-linear color corrections. 

Any comments on what the X-rite DNG generated profiles are doing?
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Peter_DL
Sr. Member
****
Offline Offline

Posts: 421


« Reply #32 on: August 17, 2011, 11:24:27 AM »
ReplyReply

I am also puzzled by the discussions of the non-linearity of ACR's Exposure.  ACR's Exposure in the positive direction is simply a straight multiply with a hard clip, just like digital camera exposure.  In the minus direction, the only difference is that ACR tries to keep clipped whites white (i.e., we did not feel it was photographically useful to let speculars turn into gray blobs, though we still let users accomplish that with the point curve)...

So in the minus direction it is really a jump function ?

Clipped white is fixed at RGB 255 while a multiplier < 1 is applied all other values, without any smoothing ?

Hmm

Peter

--
Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1291



WWW
« Reply #33 on: August 17, 2011, 11:53:45 AM »
ReplyReply

I am also puzzled by the discussions of the non-linearity of ACR's Exposure.  ACR's Exposure in the positive direction is simply a straight multiply with a hard clip, just like digital camera exposure.

That is why I don't understand why it appears to be a non-linearity near zero. These are the curves that model a change in Exposure into ACR with respect to the 0.0 setting (I already showed them to you in Dpreview). I just compared (input, output) pairs, I didn't any gamma calculation. And the ouput profile used was Adobe RGB, which is supposed to be a standard gamma profile (strictly applying f^(1/g) along the whole range).

So, what are those non-linearities near 0?



It appears to me that ACR is using a non-standard gamma 2.2 for Adobe RGB, something like sRGB's gamma.

Regards
Logged

N Walker
Sr. Member
****
Offline Offline

Posts: 300


WWW
« Reply #34 on: August 17, 2011, 12:27:51 PM »
ReplyReply

Information for the histogram and RGB readout, Lightroom uses tonal response curve similar to sRGB.
« Last Edit: August 17, 2011, 01:10:47 PM by Nick Walker » Logged

bjanes
Sr. Member
****
Online Online

Posts: 2825



« Reply #35 on: August 17, 2011, 01:31:45 PM »
ReplyReply

That is why I don't understand why it appears to be a non-linearity near zero. These are the curves that model a change in Exposure into ACR with respect to the 0.0 setting (I already showed them to you in Dpreview). I just compared (input, output) pairs, I didn't any gamma calculation. And the ouput profile used was Adobe RGB, which is supposed to be a standard gamma profile (strictly applying f^(1/g) along the whole range).

So, what are those non-linearities near 0?



It appears to me that ACR is using a non-standard gamma 2.2 for Adobe RGB, something like sRGB's gamma.

Regards


Guillermo,

Adobe RGB has no linear segment near luminances of zero, but the spec allows a slope limit of 1/32 for 8 bit digital images in the range of 1-14. See Annex C of the Adobe RGB spec. Eric Chan has confirmed that this linear segment is used in all Adobe applications.

Regards,

Bill
Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1291



WWW
« Reply #36 on: August 17, 2011, 02:46:13 PM »
ReplyReply

Adobe RGB has no linear segment near luminances of zero, but the spec allows a slope limit of 1/32 for 8 bit digital images in the range of 1-14. See Annex C of the Adobe RGB spec. Eric Chan has confirmed that this linear segment is used in all Adobe applications.

Thanks Bill. That must be the reason for the non-linearity in the calculated curves (which of course doesn't mean any non-linearity in the way Exposure works in ACR).

So if the gamma slope in Adobe RGB's implementation is limited (i.e. Adobe RGB's gamma lifts the shadows less than a pure gamma), and we open an image encoded with a pure 2.2 gamma assigning it to Adobe RGB, we will be seeing the deep shadows a bit brighter than expected. My concern about this is because I use pure 2.2 gamma images and then assign them in Adobe RGB, but I find OK to have some shadow lifting.

Regards
Logged

sandymc
Sr. Member
****
Offline Offline

Posts: 270


« Reply #37 on: August 18, 2011, 01:38:37 AM »
ReplyReply

We actually discourage (but do not prevent) the use of 3D tables in the HueSatMap* tags for the reasons you mention.  The preferred use of the HueSatMap* tags is a 2.5D table, indexed by hue & sat only, to perform non-linear color corrections.  This can be important with cameras with suboptimal spectral sensitivities or when photographing under difficult illuminants.  Lightness-dependent corrections are not necessary because the input data is in a linear, scene-referred space.  The Chart Wizard feature of the DNG Profile Editor writes into the HueSatMap* tags and writes only 2.5D tables.

Eric,

Interesting. When I did testing on ETTR about two years ago, using the LR 2.4 (which was current then), there were consistent hue shifts creeping in from somewhere, just by adjusting exposure, significantly more so with LR than with e.g., Capture 1. Mostly in blue. I assumed at the time that it was a combination of profile and tone curve. And no, no channels were saturating. So, if the profile wasn't guilty, the question becomes what was Huh

Regards,

Sandy
Logged
sandymc
Sr. Member
****
Offline Offline

Posts: 270


« Reply #38 on: August 18, 2011, 08:11:32 AM »
ReplyReply

Eric,

Ok, so I can't resist a good (technical) puzzle, so I wasted a few hours trying to track this down. What I did was to create two synthetic DNG images of a GM 24 chart, using an M9 file as template, one at nominal exposure, the other at -2EV, and looked at them in LR 3.4.1

The results were strange:
  • With nominal exposure, the CC11 patch is 72.1, 83.7, 31.0, with default settings for black, brightness and contrast
  • With -2EV, compensated by +2EV on the exposure control, the patch is 70.6, 82.6, 25.7 also with defaults, so a hue change in the blue
  • With nominal exposure, but black, brightness, contrast all at zero, the patch is 55.3, 64.5, 22.7
  • with -2EV, but black, brightness, contrast all at zero, the patch is 54.2, 63.6, 22.7, so no significant hue shift
  • Setting each of black, contrast, brightness individually to their defaults shows no major hue shift
  • With -2EV, compensated by +2.35EV on the exposure control and -34 on the contrast control so as to match the luminance of the top and bottom grey patches more precisely than the +2EV did, the patch goes to 72, 83.6, 36.1. So the red and green match nominal exposure precisely, but blue doesn't

This is the same for Adobe Standard and the embedded profile - a hue change in the blue channel, which also is the same as I got two years ago with real images from a Canon. It's a small change, hardly visible, but it is there.

Given the exact match when the exposure controls are at zero, I don't think that this can be an issue with the test images, or saturation in any channel. So I can only conclude that LR does have some kind of a hue shift when its exposure controls interact...... Shocked

Happy to send you the test images if you want.

Regards,

Sandy

Logged
ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #39 on: August 18, 2011, 10:00:08 AM »
ReplyReply

Tone controls such as black point (especially), brightness, and contrast work on individual RGB channels so will cause hue shifts.  I would only expect the exposure slider to behave as a true gain in the resulting output with black point set to zero, brightness and contrast neutral, and the tone curve set to linear.
Logged

emil
Pages: « 1 [2] 3 4 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad