Expose to the Right, even in dark scenes?

[mitchell]

Advertisement (gone after registration)

In many shots, I seem to get richer color by not exposing to the right. This is especially true when the scene itself is dark as in an evening shot.

 

But, I guess the "correct" way to deal with this is to expose to the right, and pull the exposure down in post processing. This should give better transitions, etc.

 

Is this what others think?

 

Thanks for any help.

 

Mitchell

[Guest WPalank]

Or keep exposing to the right, keep exposure where it is and push the black point to the right (doesn't take much) to get richer colors.

[hoppyman]

In many shots, I seem to get richer color by not exposing to the right. This is especially true when the scene itself is dark as in an evening shot.

 

But, I guess the "correct" way to deal with this is to expose to the right, and pull the exposure down in post processing. This should give better transitions, etc.

 

Is this what others think?

 

Thanks for any help.

 

Mitchell

Mitchell, essentially the rationale for exposing to the right is use the largest practical amount of the available dynamic range. More tonal values to start from. Those tones then get non-linear re-distribution. So if you like , finer gradation which is especially helpful in the darks (that have many fewer tones than the lights) . Having captured that larger range of tones, then you adjusting your exposure (STARTING WITH MORE VALUES) to the levels you want.

ETR is not the only successful technique of course.

However with any exposure underexposure is harmful to the darks.

Someone else mentioned, adjust your exposure as you want it than try moving your black point in (clipping)

Et voilà! much noise disappears and you haven't thrown away up to half of your dynamic range by not capturing the first (brightest) stop

[sandymc]

For those interested in a longer read, I published a new blog post on expose to the right, and why I believe its pretty much always wrong, here: ChromaSoft: Why "Expose to the Right" is just plain wrong

 

Sandy

[hoppyman]

For those interested in a longer read, I published a new blog post on expose to the right, and why I believe its pretty much always wrong, here: ChromaSoft: Why "Expose to the Right" is just plain wrong

 

Sandy

Hi Sandy I have the utmost respect for your expertise. I don't follow some of your methodology here. This is not to cast doubt on what you have shared in your blog (or your qualifications). I want to better understand your conclusions and methods.

 

Perhaps I have a different concept of the Expose To The Right process. The comment attributed to Thomas Knoll of course is the basis of my views, as a user without technical qualifications on this.

I don't understand the "underexpose ETTR" idea. When you are talking about 'driving the shadows down to the left' do you mean adjustment (tonal redistribution) in post processing? Are you referring to some technique for measuring different portions of the scene as a part of making an exposure calculation (like zone exposure for example)? I just want to place those highlights then the rest falls below that. High or low contrast, just capture the maximum number of tonal values to start from.

I guess that makes me a 'classical' proponent?

 

I don't follow the rationale for changing the ISO as the exposure alteration mechanism. I'm fine with your explanation of the sensor noise ratio comparison and certainly the parameter of avoiding any clipping. But why alter the exposure by changing the ISO? This appears to relate to ISO noise comparison more than how many tonal values are captured?

 

Can you expand on your exposure determinations to avoid saturation? Converting into what colour space for example? I guess for the purposes of your test it isn't critical providing you are consistent, I just think of the indicated clipping (on the camera LCD) as a guide with limitations as part of determining the exposure. That is to say, not assume that the LCD indication is gospel and underexpose to avoid any possible clipping indication.

I follow the effect of the in camera noise reduction in your examples. Can you comment on the applications with the M8/M9? Candidly I am only interested in the practical applications with those specific cameras!

 

You've said in your conclusions that the compression in the M8 has not been proven to be a practical disadvantage. That's fine, I certainly can't provide contrary evidence. Perhaps as the M9 gets more widely distributed that can be conveniently put to the inquisition with its compression options.

 

I'm learning a lot from your hue twist references and am looking foward to M9 profiles appearing in ACR &LR naturally. Again only looking from an M-centric point of view with basically the Adobe options of Standard and then my own profiling.

Thanks again for your contributions on all of this.

[sandymc]

I don't understand the "underexpose ETTR" idea. When you are talking about 'driving the shadows down to the left' do you mean adjustment (tonal redistribution) in post processing? Are you referring to some technique for measuring different portions of the scene as a part of making an exposure calculation (like zone exposure for example)? I just want to place those highlights then the rest falls below that. High or low contrast, just capture the maximum number of tonal values to start from.

I guess that makes me a 'classical' proponent?

 

That does indeed make you a classical proponent!. As I noted in the article, there are a lot of different version of what ETTR should and should not be. I'm using the term ETTR to cover anything that involves setting exposure by placing the highlights on the right of the histogram, without looking at where other parts of the image will end up. That covers either overexposing a low contrast scene to move the histogram to the right, or underexposing a high contrast scene to make the highlights move from blown (off the histogram) to the right of the histogram.

 

 

I don't follow the rationale for changing the ISO as the exposure alteration mechanism. I'm fine with your explanation of the sensor noise ratio comparison and certainly the parameter of avoiding any clipping. But why alter the exposure by changing the ISO? This appears to relate to ISO noise comparison more than how many tonal values are captured?

 

If you adjust ISO down and expose normally rather than overexposing at higher ISO, you get the same noise performance, but because you didn't overexpose, you get problems with color shifts, etc in post.

 

Can you expand on your exposure determinations to avoid saturation? Converting into what colour space for example? I guess for the purposes of your test it isn't critical providing you are consistent, I just think of the indicated clipping (on the camera LCD) as a guide with limitations as part of determining the exposure. That is to say, not assume that the LCD indication is gospel and underexpose to avoid any possible clipping indication.

 

Seeing as I had lots of time to do this particular exercise, I had the luxury of looking at the histograms of the actual raw data on a PC, and adjusting to ensure there was no saturation. From a LCD display on the back of a camera, I completely agree with you that its a problem. My experience is that you have to allow for a quite a margin - up to stop - for variations between what you see on a camera histogram and reality.

 

 

I follow the effect of the in camera noise reduction in your examples. Can you comment on the applications with the M8/M9? Candidly I am only interested in the practical applications with those specific cameras

 

I think the M8/M9 are the same as any other camera in this regard - don't use ETTR unless you're at the lowest ISO setting.

 

You've said in your conclusions that the compression in the M8 has not been proven to be a practical disadvantage. That's fine, I certainly can't provide contrary evidence. Perhaps as the M9 gets more widely distributed that can be conveniently put to the inquisition with its compression options.

 

Well, I was careful to say that the M8 compression hasn't been an issue under normal conditions. If you need to post process aggressively e.g., recovering a badly exposed shot, etc, then the compression can be an issue. But given that ETTR is really only a technique you can use if you have the time to get exposures right, my argument is that you're better off just exposing correctly.

 

The M9's compression is a bit of different issue; as I've mentioned on other threads I don't like the way Leica have implemented it. While it uses the same compression curve as the M8, because the M9 has non-zero black levels, at high ISO you lose to lot of codes, and hence tonality to "lower than black" data; up to 20% of the camera available levels. I think it's possible that that will be an issue even with normal exposures; to me, and I think Jamie has mentioned this as well, some compressed images look just a little "wrong".

 

Sandy

[hoppyman]

Advertisement (gone after registration)

Thank you Sandy. That made some issues clearer and introduced some new ones for me to think about:)

[Jamie Roberts]

Sandy, I agree with you on ETTR fanaticism. You do get hue shifts at high ISOs when you do this, and it's just not necessary with lower ISOs.

 

But correct exposure is subject related, so perhaps we should also mention the absolute noise floor of a camera at high ISOs as a limiting factor.

 

I mean, doesn't your position need to be modified to take into account the need to be intelligent about normal exposure in extreme conditions (which is where this always comes up) You can't just let significant values "fall off the cliff" when higher ISOs are called for...and in that sense you're "exposing to the right" and forgetting the histo entirely.

 

Another way to think of this is that your test is all evenly illuminated; normal exposures aren't, and sometimes trade-offs with hue shifts are completely ok if you gain shadow detail and get the shot, right?

 

So here's a scenario: I have an M8 and I want to expose a dark scene with some shadow detail in the lower quartertones. It's a high contrast (night time) scene and I don't care about speculars or even about other highlights much... all the content of the image is in the shadows.

 

I can't use a low ISO (here meaning 640) because "correct" exposure for the shadow detail means I'd be shooting 1/20th and let's say I can't hand-hold that reliably and I have no tripod.

 

But ideally, I'd really like to expose at 640 even a bit more slowly (say 1/10th) if I could, just so I could ensure tonality in the shadows.

 

But I can't do that--and dropping the ISO, while making noise better, would only make the shutter speed situation worse.

 

So what should I do? If I switch to ISO 1250 I will gain noise due to the amplification of the image. What's worse, I only have one more stop of detail left now and need to nail the shadows at a neutral level to have a hope of luminance information, right?

 

So I have a quandry: I can't hand-hold the shot at lower ISOs and I don't have any wiggle room at the higher ISO. ISO 2500 means I have even less room for error...

 

But a higher ISO means I have choices around my shutter selection. So my personal choice is to push the exposure at higher ISOs in this case, because I increase my chances of getting a usable / printable result that way.

 

In this case, for example, I'd even go to ISO 2500 if I could maintain the shutter and get more "levels" of exposure if the shadows are that important. I can even bracket at that level: normal exposure here is now 1/80th but I can hand-hold down to 1/40th if I need to...

 

So I'd rather deal with inherent noise around the middle-gray exposure (and bring them back / make them darker in post) than having the shadows drop off entirely and not being able to print them.

 

Am I wrong about this? What are your thoughts here?

[sandymc]

Jamie,

 

I'd agree with that.

 

What you're describing is a situation where as a photographer you're taking a decision about (a) what parts of the image are most important to you and ( what the least of several evils are, and deciding the exposure to get the best image you can. I have no problem at all with that.

 

Where I have problems is the belief that applying ETTR without thinking through what you're doing, and what you need to achieve, will always give you the best possible image you can get.

 

Regards,

 

Sandy

[Jamie Roberts]

Thanks Sandy--just wanted to confirm I wasn't missing something!