Expose to right, it is as simple as

[MarkM]

That does not characterise a low-contrast scene. A limited range of luminance values does. So take all those colours from your triple-100-to-triple-155 sub-cube—and then add all the other RGB values that have the same luminance values as any of the points inside your sub-cube. You'll end up with many more RGB values that definitely do not fit ithe triple-200-to-triple-255 sub-cube.

You are trying to have it both ways—you keep talking about luminance when you are making an argument using RGB geometry. There is no luminance axis in this model.

What you seem to be saying is that you want to talk about colors that are not with this 100, 155 cube. I think that's what you are saying when you say "other RGB values that have the same luminance." So take [155, 155, 155] which has a L* value of 65 (in ARGB) and consider other colors which have the same luminance but are more saturated—say a LAB value of (65, 20, 0). This color is no longer in the sub cube, but will be something like [179, 143, 157]. If you include this color in the scene, your sub-cube just got a little bigger and more rectangular, but still you can push the values up until the red channel starts to clip and you'll still have exactly the same number of discreet values. You can't give me any set of RGB numbers that I can't add a constant to and end up with an isomorphic set so long as I don't let the channels clip.

[joofa]

Why should that happen?. Changing exposure is just a linear scaling of the captured RGB triplet, RGB={400,200,800} overexposed by 2 stops becomes {1600,800,3200}. If no clipping occurs, how can you say exposing more reduces colour gamut? it will simply have more resolution in defining the colours, and less noise, but not less colour definition.

I think what 01af is saying is that the some colors in volume of colors that are close by in luminance might get out of gamut / clipped if exposure is increased.

Joofa

[01af]

You are trying to have it both ways—you keep talking about luminance when you are making an argument using RGB geometry.

Strange how people falsely believe RGB triples didn't contain any luminance information ...  

You can't give me any set of RGB numbers that I can't add a constant to and end up with an isomorphic set so long as I don't let the channels clip.

Of course I can't (as long as you include zero to the possible range of constants). But the maximum value of the constant you can add largely depends on the saturation of the most-saturated colour that you want to keep unclipped. That's my point ... it may seem trivial in the context of this disciussion ("do not clip any of the raw channels") but isn't out in the real-life field.

[Peter_DL]

So please understand: Pushing up the exposure reduces noise, yes, but it also reduces the colour gamut.

Maybe for example:
D65 white like with sRGB has no real AbsCol representation in a D50 space like ProPhoto RGB.
Channel clipping occurs (and vice versa). At the very top of the gamut even large ProPhoto RGB gets so tiny that a slightly different shade of white can't be held. At least not without reducing exposure / linear downscaling.

No idea though if this would be of any practical (ETTR) relevance here.

Peter

--

[Guillermo Luijk]

I think what 01af is saying is that the some colors in volume of colors that are close by in luminance might get out of gamut / clipped if exposure is increased.

So 01af is saying that clipping may occur if we expose too much? such a discovery deserves extra experimentation...

[joofa]

Maybe for example:
D65 white like with sRGB has no real AbsCol representation in a D50 space like ProPhoto RGB.
Channel clipping occurs (and vice versa). At the very top of the gamut even large ProPhoto RGB gets so tiny that a slightly different shade of white can't be held. At least not without reducing exposure / linear downscaling.

No idea though if this would be of any practical (ETTR) relevance here.

Peter

--

But, the clipping direction doesn't have to be towards the top near white. It can be sideways.

Joofa

[joofa]

So 01af is saying that clipping may occur if we expose too much? such a discovery deserves extra experimentation...

No, if I have understood him correctly, a simple visualization in 2D or 3D might suffice.

Joofa

[MarkM]

But the maximum value of the constant you can add largely depends on the saturation of the most-saturated colour that you want to keep unclipped.

In other words, expose to the right without clipping color channels. Funny, I thought that's what everyone has been saying the whole time.

[joofa]

In other words, expose to the right without clipping color channels. Funny, I thought that's what everyone has been saying the whole time.

No, if I understood him correctly, the clipping he is talking about is in a standardized space, say sRGB, and not camera channel clipping.

Sincerely,

Joofa

[Peter_DL]

So 01af is saying that clipping may occur if we expose too much? such a discovery deserves extra experimentation...

Valid point.
The argument that a gamut gets smaller at the top
is annulled by the ETTR constraint to avoid channel clipping.

Peter

--

[Guillermo Luijk]

if I understood him correctly, the clipping he is talking about is in a standardized space, say sRGB, and not camera channel clipping.

The argument that a gamut gets smaller at the top
is annulled by the ETTR constraint to avoid channel clipping.

ETTR is about capturing RAW data, just photons being counted by the sensor and producing RGB numbers for which exposure is just a scaling factor. As long as no RAW clipping occurs, exposure in the capture has zero effect on any later colour space transformations.

[joofa]

Valid point.
The argument that a gamut gets smaller at the top
is annulled by the ETTR constraint to avoid channel clipping.

Peter

--

I think what he is saying is that if you expose too much then some colors might get clipped in a standardized RGB space even if camera channels are not clipped.

Joofa

[Peter_DL]

ETTR is about capturing RAW data, just photons being counted by the sensor and producing RGB numbers for which exposure is just a scaling factor. As long as no RAW clipping occurs, exposure in the capture has zero effect on any later colour space transformations.

So a Raw file would have no color space ?

Have fun,
and a nice evening.

Peter

--

[Guillermo Luijk]

I think what he is saying is that if you expose too much then some colors might get clipped in a standardized RGB space even if camera channels are not clipped.

So a Raw file would have no color space ?

hehe it could, it could not (who cares?), the fact is that exposure in the capture (which is the only place where discussing about ETTR makes sense), has no consequences in later colour conversions since it can be freely adjusted. Reducing exposure by 2 stops in the RAW data just consists of multiplying all RAW numbers by 0,25.
 

[Peter_DL]

... the fact is that exposure in the capture (which is the only place where discussing about ETTR makes sense), has no consequences in later colour conversions since it can be freely adjusted.

But the photons fist have to pass through the Bayer filters,
which finally define the "Raw gamut" including mentioned tininess at the top.

--

[joofa]

Reducing exposure by 2 stops in the RAW data just consists of multiplying all RAW numbers by 0,25.
 

Yes, true. But, it is a user's prerogative what exposure they want to develop for.

Joofa

[ejmartin]

But the photons fist have to pass through the Bayer filters,
which finally define the "Raw gamut" including mentioned tininess at the top.

--

Guillermo has it right, so long as there is no data loss in capture and the converter treats the data faithfully, nothing is lost in ETTR.  ETTR pulled maps to the same output range as a non-ETTR exposure since the pulling is done long before the map to the output color space and any issues associated with colors mapping outside the output gamut.  The output routines are dealing with the same data apart from the ETTR data having better S/N.

[madmanchan]

Clipping in a standard space like sRGB or Adobe RGB is a completely separate matter from ETTR. Clipping and gamut issues have to do with the gamut mapping algorithm, not the set of colors the camera can record. The main thing that affects the recorded set of colors is the optical path, namely the spectral transmission of the lens and the spectral sensitivities of the sensor.

You can take pictures of a vivid rose whose tristimulus valued are outside standard color spaces, regardless of what exposure you use.

[joofa]

You can take pictures of a vivid rose whose tristimulus valued are outside standard color spaces, regardless of what exposure you use.

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.

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.

Sincerely,

Joofa

[bjanes]

Clipping in a standard space like sRGB or Adobe RGB is a completely separate matter from ETTR. Clipping and gamut issues have to do with the gamut mapping algorithm, not the set of colors the camera can record. The main thing that affects the recorded set of colors is the optical path, namely the spectral transmission of the lens and the spectral sensitivities of the sensor.

You can take pictures of a vivid rose whose tristimulus valued are outside standard color spaces, regardless of what exposure you use.

That is an interesting point. I was under the impression that ProPhotoRGB is able to represent all of the color values captured by the camera. In my own experiments with ETTR, I have encountered clipping with ETTR while rendering into ProPhotoRGB, but this can be removed by decreasing the exposure in ACR. This is not surprising since RGB color space gamuts decrease with luminance. One can also decrease clipping when rendering into sRGB by decreasing the exposure, but it is better to use a wider color space. The critical question with ProPhotoRGB is what exposure value would produce a correct representation of the image.

Regards,

Bill