Expose to right, it is as simple as

[Guillermo Luijk]

I think the issue is that one takes the picture of a rose at some exposure with no clipping when converted to some RGB, and then one increases exposure optically (with no exposure compensation in software) and sees clipping. If that can really happen, and if that is what 01af is saying, then I think it is important that one should be aware of it.

I confirm you that can really happen, it is called 'RAW clipping'. And ETTR is about preventing it.

What Gullermo and Emil are saying is fine, but they are operating on a preconceived notion of doing exposure adjustment in software to arrive at a predetermined result. There is no rule in the world that suggests what exposure adjustment to do in software. A user is free to do no exposure adjustment, or worse increase exposure for development, with the possible caveat of clipping occuring.

What Emil and me are saying is just that what 01af said is not correct: ETTR does not reduce any colour separation (or whatever he wants to call it).

Your exposure adjustment argument is a fallacy since software exposure adjustment is intrinsic to ETTR, the topic under discussion. Besides that, unless you like greenish images, all RAW files (ETTR'ed and non ETTR'ed) need exposure adjustment performed by the software, it is called white balance.

Regards

[madmanchan]

Bill, when "rendering into" ProPhoto or any other standard colorimetric space, the white balance step is applied and this step can result in clipping (since scale factors are applied to the raw data). For example, a perfectly valid raw red-channel value of a rose at 0.8 (not clipped) may get mapped to 2.0 during white balance, then clipped to the representable output range of [0,1]. Reducing the exposure in software effectively reduces the digital scale factors, and thereby avoids the clipping (*).

So, it is true that the initial rendering of a raw image as seen in a raw converter may appear to have clipped colors when using ETTR, but as long as the original raw channels were not clipped (in the image areas you care about), then you still have all the image data preserved which you can then use to color-map and tone-map as desired (curves, local adjustments, etc.), without loss of detail.  

Eric


(*) Of course, reducing the exposure not only removes the clipping, but it also makes the overall image darker, which may or may not be what you really want.  That's where tone mapping and really "developing" the picture to taste come into play (separate issue, of course).  

[joofa]

I confirm you that can really happen, it is called 'RAW clipping'. And ETTR is about preventing it.

I think I mentioned we are talking about RGB space clipping and not raw camera channel clipping.

As Eric Chan mentioned there are ways to avoid RGB clipping, at least in some situations if not all. The sole exercise is knowing the options and possible caveats in increasing optical exposure and possible RGB clipping. Not what you think is the definition of ETTR, or why everybody should adhere to the workflow you want to achieve a preconceived result.

Sincerely,

Joofa

[Peter_DL]

What Emil and me are saying ...

Seems we are talking about different subjects.

One question was, if upon increasing exposure, a real-world color could be pushed out of a camera’s gamut. Camera profiles – as can be studied in 3D glory for example here –may look like big matrix spaces, however, they are not endless and get tinier at the top.

My initial assumption was that such out-of-camera-gamut capture would go along / being prevented by Raw channel clipping. Not sure. The alternative would be to receive (unclipped) Raw RGB data which are not unambiguously assigned / interpretable by the corresponding camera profile.

Peter

--

[Guillermo Luijk]

Not what you think is the definition of ETTR, or why everybody should adhere to the workflow you want to achieve a preconceived result.

The workflow of ETTR is clear: expose as much as possible in the camera without RAW clipping, and bring exposure back in post processing. There is no link to RGB spaces clipping in the context of ETTR.

Of course you are free to reformulate 'ETTR' in a creative way introducing colour space theory, but that will be a different thing to what this thread is about.

My initial assumption was that such out-of-camera-gamut capture would go along / being prevented by Raw channel clipping. Not sure. The alternative would be to receive (unclipped) Raw RGB data which are not unambiguously assigned / interpretable by the corresponding camera profile.

ETTR doesn't impose any limitation regarding possible gamut clipping after RAW conversion. As long as you don't clip RAW data, everything you can achieve with an ETTR'ed capture is the same that you can achieve with any lower exposure capture, with the difference of having less visible noise.

Regards

[joofa]

Of course you are free to reformulate 'ETTR' in a creative way introducing colour space theory, but that will be a different thing to what this thread is about.

Regards

I like it when people want to define what the thread is about, or how it should proceed, or should be interpretted - kind of like my way or the highway. This discussion started from the concerns of 01af regarding the implications of luminance on RGB clipping, so of course color theory is inolved.

Sincerely,

Joofa

[kwalsh]

I think the issue is that one takes the picture of a rose at some exposure with no clipping when converted to some RGB, and then one increases exposure optically (with no exposure compensation in software) and sees clipping.

Of course, but that isn't ETTR - that is just over exposing.  It is a well known characteristic and is the basis for muted colors in high-key images (or low-key images for underexposing).  Nothing new here, and nothing related to ETTR.

If that can really happen, and if that is what 01af is saying, then I think it is important that one should be aware of it.

If that is what 01af is saying then he is correct, and this has been known well before digital.  I can't imagine anyone not being aware of it already.  Almost all practical output color spaces have the most saturated colors near middle gray.  If you want saturated colors it is best to target those colors for a mid-gray exposure in the output.  People did that with Kodachrome for ages and ages.  Very old news.

This isn't relevant at all to ETTR assuming your ETTR exposure doesn't clip any RAW color channels (outside of specular highlights).  What it is relevant to is your output color space which occurs after exposure adjustments have been made to the RAW data.  When exposure compensation is applied in the RAW converter the output tones will be back in the widest part of the gamut.  If there was no channel clipping these exposure compensated channel levels will be identical to those of a "normal" exposure and the output gamut at that tonal level will be the same.  There will just be lower photon shot noise in the ETTR exposure, the whole point of ETTR.

What Gullermo and Emil are saying is fine, but they are operating on a preconceived notion of doing exposure adjustment in software to arrive at a predetermined result.

But that is the notion of ETTR - expose above your intended output exposure if there is headroom to spare.  If you aren't doing exposure adjustment in software then you aren't doing ETTR, you are just following the camera metering.

There is no rule in the world that suggests what exposure adjustment to do in software. A user is free to do no exposure adjustment, or worse increase exposure for development, with the possible caveat of clipping occuring.

Of course, again high-key and low-key photography.  It isn't limited to software at all either.  This decision is made in camera with transparencies, at printing time with negatives, in PP or in camera with JPEGs, and ideally in the RAW converter with RAW images.

I just don't see how any of 01af's discussion or your points here are relevant to ETTR.  We are now talking about color representation in the processed photo after artistic decisions.  That has little if anything to do with ETTR.  ETTR is simply getting the lowest noise photon counts without introducing non-linearity (clipping) for any relevant scene detail.  It is a fairly mechanical process.  The question of output exposure comes afterwords and is independent of whether the exposure was "normal" or ETTR.

af01 hasn't been very clear about what issue he is addressing.  If it is about output color spaces and the limited gamut width in the highlights and shadows, well then "duh", I don't think anyone is going to consider that news.  If he is claiming that an ETTR exposure without clipped RAW channels has a restricted gamut compared to a lower exposure after both have been properly exposure compensated in the RAW converter then he is flat out wrong - plain and simple.

Ken

[joofa]

I just don't see how any of 01af's discussion or your points here are relevant to ETTR.  

I have tried not to define what exactly is ETTR or what is relevant to ETTR. I have just tried to present what I thought 01af meant as it seems almost everybody could not figure out what he was trying to say - most here were mixing raw clipping with RGB clipping, when I thought that 01af meant RGB clipping. May be he means something totally different.

Sincerely,

Joofa

[bjanes]

One question was, if upon increasing exposure, a real-world color could be pushed out of a camera’s gamut. Camera profiles – as can be studied in 3D glory for example here –may look like big matrix spaces, however, they are not endless and get tinier at the top.

My initial assumption was that such out-of-camera-gamut capture would go along / being prevented by Raw channel clipping. Not sure. The alternative would be to receive (unclipped) Raw RGB data which are not unambiguously assigned / interpretable by the corresponding camera profile
--

It may seem pedantic, but digital cameras and other input devices do not have a gamut, as strictly defined (see the Munsell FAQ). Therefore, it would not be possible to push a color beyond the "gamut" of the camera. However, digital sensors do clip. I am not sure that this clarifies the topic under discussion, but it is an interesting fine point.

Regards,

Bill

[Eric Myrvaagnes]

I like it when people want to define what the thread is about, or how it should proceed, or should be interpretted - kind of like my way or the highway. This discussion started from the concerns of 01af regarding the implications of luminance on RGB clipping, so of course color theory is inolved.

Sincerely,

Joofa

This thread was started by ErikKaffehr, not by O1af, who didn't pipe in until reply # 142. Go back to Erik's original post and you will see that the thread did indeed start with Erik's comments about ETTR, and NOT about O1af's concerns about luminance. Nevertheless, O1af's ambiguous suggestions seem to me to have been answered pretty effectively by several knowledgable responders/

[kwalsh]

I have tried not to define what exactly is ETTR or what is relevant to ETTR. I have just tried to present what I thought 01af meant

Ah, got it.  Thanks!  Maybe he'll swing by again and let us know.

Cheers,

Ken

[crames]

If that is what 01af is saying then he is correct, and this has been known well before digital.  I can't imagine anyone not being aware of it already.  Almost all practical output color spaces have the most saturated colors near middle gray.  If you want saturated colors it is best to target those colors for a mid-gray exposure in the output.  People did that with Kodachrome for ages and ages.  Very old news.

It depends on how you look at it. If you view the RGB cube in a perceptual space, for example CIELAB, the maximum available saturation does not necessarily correspond to the mid-gray. This example shows that in AdobeRGB, the maximum chroma is nearest to the lightness of white for yellow, green, and cyan.



(image created with RGB Cube)

[bjanes]

It depends on how you look at it. If you view the RGB cube in a perceptual space, for example CIELAB, the maximum available saturation does not necessarily correspond to the mid-gray. This example shows that in AdobeRGB, the maximum chroma is nearest to the lightness of white for yellow, green, and cyan.

That is a good point, but it is difficult to read the luminance at maximum chroma from your 3-D graph. A better view can be obtained in ColorThink Pro using the slicer. The maximum chroma for cyan occurs at a L* of about 87 as shown below. Nonetheless, a RGB space such as AdobeRGB does have decreased chroma as L* approaches 100 as shown in the second image, even though maximal chroma does not occur at L* of 50. As previously pointed out, one can place the luminance as desired when rendering a raw ETTR into aRGB and it is not necessary to expose for mid-gray. Chroma clipping can be reduced by rendering into ProPhotoRGB.

Regards,

Bill

[crames]

That is a good point, but it is difficult to read the luminance at maximum chroma from your 3-D graph. A better view can be obtained in ColorThink Pro using the slicer. The maximum chroma for cyan occurs at a L* of about 87 as shown below. Nonetheless, a RGB space such as AdobeRGB does have decreased chroma as L* approaches 100 as shown in the second image, even though maximal chroma does not occur at L* of 50. As previously pointed out, one can place the luminance as desired when rendering a raw ETTR into aRGB and it is not necessary to expose for mid-gray. Chroma clipping can be reduced by rendering into ProPhotoRGB.

AdobeRGB maximum chroma yellow (255,255,0) is at L* = 97. Max blue chroma (0,0,255) is at L* = 33. I think it is safe to say that one can't generalize about which luminance or lightness will yield the strongest colors. It depends on both the particular color, and the color space.

I agree with you and the others who say that, as long as there is no clipping, ETTR can be used without losing any colors in the capture. Has anyone produced a demonstration that proves otherwise?

On the output side, there is a related problem reproducing scenes with highly-chromatic white points. Here is an interesting paper by Heckaman & Fairchild, "Expanding display color gamut beyond the spectrum locus."  View the example images full screen in a dark room - very effective! They call it “pushing down the white point of the display." Around here it might be called Display To The Middle. 

[kwalsh]

It depends on how you look at it. If you view the RGB cube in a perceptual space, for example CIELAB, the maximum available saturation does not necessarily correspond to the mid-gray. This example shows that in AdobeRGB, the maximum chroma is nearest to the lightness of white for yellow, green, and cyan.

Yes, thank you for correcting my oversimplification.  It really depends on the chromaticities of the primaries of the color space.

Ken

[hjulenissen]

Second, ETTR will optimize separation of tones but sacrifice separation of colours. So if the subject's dynamic range is small then don't push it up all the way to the right. Instead, leave it centered ... or push it up half-way between center and right.
...
For the optimal exposure, you'd need to balance dynamic range versus colour range.

So after a couple of pages of discussion, it seems to me that 01AF thinks that ETTR means undesirabe clipping (either in camera or in raw development), while most others think that is generally wrong, and everyone agress that clipping is bad for colors?

While having example images showing what one tries to argue is always a good thing, I think that what is really needed is a common understanding of what ETTR really means. If everyone use their own definition of ETTR, then discussing it merits (or lack thereof) is pointless.

-h

[stamper]

And if someone, who isn't as knowledgeable and is trying to learn and hoping for a consensus of opinion, is lurking then it is disappointing, to say the least. What probably makes it worse is that there are one or two stirring the pot and obscuring any real information. Then again I and others have been told that we don't need to read if we don't have to?

[hjulenissen]

And if someone, who isn't as knowledgeable and is trying to learn and hoping for a consensus of opinion, is lurking then it is disappointing, to say the least. What probably makes it worse is that there are one or two stirring the pot and obscuring any real information. Then again I and others have been told that we don't need to read if we don't have to?

What are you saying?

Total consensus of opinion is very rare. This is the case for academia, politics and internet forums. Part of the skill-set needed to learn from these sites is being able to distinguish relevant, knowledgeable posts from posts that are less so, and to repeat important tests for yourself when needed.

01af have been asked to provide images /test-procedures that supports his claim. When he/she does so, it will be a lot easier for us to understand what he/she really means, and how relevant it is to our own line of photography.

-h

[Peter_DL]

It may seem pedantic, but digital cameras and other input devices do not have a gamut, as strictly defined (see the Munsell FAQ). Therefore, it would not be possible to push a color beyond the "gamut" of the camera. However, digital sensors do clip. I am not sure that this clarifies the topic under discussion, but it is an interesting fine point.

Yes, that’s an quite academic definition of "color gamut"
which also seems to consider that a camera is supposed to react on any light source such as e.g. a spectrally pure laser beam, even though such stimulus may lie outside of what we draw as a camera color space.

Anyway, the final conclusion that it doesn’t matter here for ETTR sounds logical,
and Guillermo is of course right when saying that any +Eposure can be undone by linear down-scaling of the (unclipped) Raw RGB data.

Could still be interesting to hear the "defense" of Andrey Tverdokhleb and Iliah Borg
who brought up this thesis about "ETTR is very harmful for colors - midpoint is the most colorful place in A900 gamut" as quoted here.

Best regards, Peter

--

[fdisilvestro]

Could still be interesting to hear the "defense" of Andrey Tverdokhleb and Iliah Borg
who brought up this thesis about "ETTR is very harmful for colors - midpoint is the most colorful place in A900 gamut" as quoted here.

Best regards, Peter

--

The important comment is this one:

Veiling glare from lens and sensor are the culprits here

Nothing to do with ETTR at least from the theoretical point of view. If RAW channels aren´t clipped, there will be no issues with colors unless some external factor (like veiling glare) is involved.