6
\$\begingroup\$

I read an article called Expose (to the) Right, which explained why you should try to get the graph of the histogram as much to the right of the scale as possible. The reasoning is that DSLRs record much more detail in bright areas of the subject than in darker areas.

You could then stop back in your photo editing tool to a "normal" correct exposed picture but with reduced noise and more range than if you'd done the shot with the "normal" correct exposure up front.

That all makes sense to me, but in the end the author quotes someone else stating:

For film based photography, the highlight end of the scale is compressed by the shoulder portion of the D/log E curve. So as brighter and brighter objects are photographed, the highlight detail gets gradually compressed more and more until eventually the film saturates. But up until that point, the highlight compression progresses in a gradual fashion.

Solid state sensors in digital cameras behave very differently. As light falls on a sensor, a charge either accumulates or dissipates (depending on the sensor technology). Its response is well behaved right up until the point of saturation, at which time it abruptly stops. There is no forgiveness by gradually backing off, as was the case with film.

Because of this difference, setting up the exposure using an 18% gray card (as is typically done with film) does not work so well with a digital camera. You will get better results if you set your exposure such that the whitest white in the scene comes close to, but not quite reaching, the full digital scale (255 for 8-bit capture, 65535 for 16-bit capture). Base the exposure on the highlight for a digital camera, and a mid-tone (e.g. 18% gray card) for a film camera.

Source: http://www.luminous-landscape.com/tutorials/expose-right.shtml

So … how do I do a meter reading for a DSLR if not from a gray card? My DSLR metering will always try to make my image gray, right? How can I avoid this, to make images which are "exposed to the right"?

\$\endgroup\$
1
  • \$\begingroup\$ A photographer once told me, it is camera / manufactor dependent wheter the RAW has more detail in bright areas or in dark areas. It depends on the sensor technology they use. \$\endgroup\$ May 1, 2015 at 8:30

4 Answers 4

9
\$\begingroup\$

I think the article is referring to using the histogram to judge exposure, after a test shot has been taken. Using the histogram as a guide you can increase exposure until the top of the histogram hits the right edge, indicating clipping may start happening.

If you have to rely on the in camera metering (which will meter assuming 18% reflectance as you suggest) then you can simply use exposure compensation to correct the camera exposure up a stop or two in order to expose to the right. Using the histogram is much more accurate however!

Whilst we're on the subject the luminous landscape article is a little simplistic and wrong in a few areas. Expose to the right doesn't increase detail it increases signal to noise ratio. The more light you let in, the more signal you get and hence better SNR. This even applies to increasing ISO in order to expose to the right, you increase the analogue signal above the noise floor by amplifying it.

Exposing to the right is not always desirable. Increasing SNR can come at the expense of limiting colour fidelity. The higher you go up the brightness scale the fewer colours can be represented.

Finally on every DSLR I've seen, the histogram you get is based on a jpeg image the camera creates - even if you shoot raw. I know of no camera that gives you a histogram of the raw values. This means you have to be careful of your jpeg settings (saturation and contrast) in particular when setting exposures based on the histogram.

\$\endgroup\$
4
  • \$\begingroup\$ Thanks for describing also the drawbacks and caveats of this method. \$\endgroup\$ Jul 2, 2011 at 13:01
  • \$\begingroup\$ Color fidelity does not suffer until you start saturating a channel. Now, it's true that in general the camera's histogram, even the color one, will not tell you when this happens. That is, unless you use UniWB, which is a kludge to make the histogram true to the raw data, ruining the in-camera JPEG in the process. \$\endgroup\$ Jul 3, 2011 at 13:45
  • \$\begingroup\$ Matt, you note that "The higher you go up the brightness scale the fewer colours can be represented." Isn't the LL article arguing the opposite, that (in the 12 bit example they give) that the brightest stop gets 11 bits of data, the next stop 10 bits, and so on down? \$\endgroup\$
    – mattdm
    Jul 5, 2011 at 19:58
  • \$\begingroup\$ @mattdm: Seems Matt is talking about chromaticity gamut, which is reduced when you get close to saturation. For example, in sRGB, pure blue, red and green are lost, respectively, past 7.3, 21.3 and 71.6% luminance. At 100% luminance you only have pure white: all chromaticity is lost. LL is talking about color resolution, i.e. how precisely a color can be defined, assuming it's still in-gamut. \$\endgroup\$ Jul 6, 2011 at 11:38
4
\$\begingroup\$

I'll add my 2¢ worth: at least 95% of the time, ETTR in general is a poor choice. In theory, it should be fine. It's absolutely true that increasing the signal without affecting the noise should improve the signal to noise ratio.

In fact, if (for example) you give a picture a third to half a stop less exposure, you'll still lose very little to noise (little enough that you're unlikely to notice without a direct side-by-side comparison, and it's not always obvious even then). At the same time, increasing exposure gives a much greater chance of blowing the picture completely by losing all detail in the highlights.

ETTR is roughly equivalent to a Blackjack strategy that says the optimal score is 21 so you should hit on every score up to an including 20. Doing so might increase the number of times per game you get a score of 21 -- but it almost certainly will increase the number of times you go bust and win nothing. Ultimately, it's a losing strategy.

ETTR became possible and quit making sense at exactly the same time. Essentially the only way to do ETTR is with digital photography -- but with digital photography, bracketing becomes essentially free. Instead of working hard to find the one optimal exposure at the time, you can just shoot and keep a number of exposures and pick the best one when you have plenty of time, a much larger, more accurate monitor, and the ability to measure exposure directly from the raw data instead of indirectly in a JPEG generated from the raw data.

Bottom line: if you can do ETTR, you're better off bracketing instead. If you don't have time to bracket, you almost certainly don't have time to even attempt ETTR either. If you're trying to be certain of catching decisive moments, you're much better off shooting with a slightly lower exposure to give better assurance against blown highlights, and trusting a modern camera to keep noise under control.

I also feel obliged to point out that his comments about film are a bit off as well. Ultimately, in digital photography (using ETTR or otherwise) you largely need to expose for the highlights, assuring that you place the exposure so they maintain detail, and let everything else work out from there.

Film exposure depends: with slide film, you do almost exactly the same. With color print film, you do the opposite: expose for the shadows. For B&W, the zone system lets you do some of both. You can increase contrast (at the expense of increased grain) by underexposing and over-developing, or decrease it by overexposing and under-developing. Ideally, you measure the contrast range in the subject matter, figure out how much you need to compress or expand that to get the desired range on the film, and expose and develop accordingly (and preferably use film with fine enough grain that you can use a fairly wide range of variation with impunity).

You didn't necessarily meter film based on mid-level grey exposure, and metering for digital ends up quite similar to metering for slide film.

\$\endgroup\$
14
  • \$\begingroup\$ I agree you shouldn't be trying to push the histogram to the right if you're relying on the camera's metering under changing lighting conditions, you're bound to overexpose something. But there are plenty of situations where ETTR makes sense but bracketing doesn't. If you're in the studio you can set the lights up, push the histogram to the right, and shoot away with these settings. Why in this situation would you take three shots each time - firing the flash three times as much as you need to is bad, especially if you're using battery power! \$\endgroup\$
    – Matt Grum
    Jul 5, 2011 at 9:11
  • \$\begingroup\$ Here's another example of when ETTR is better than bracketing. When shooting static scenes, e.g. landscapes, I will take an exposure, check the histogram, move it to the right, shoot again check the histogram, thus achieving ETTR, without blown highlights (as each shot is checked). This is technically not bracketing, even if you end up taking the same number of shots, and will result in much more accurate exposures than relying on the camera's metering and firing a bracketed sequence each time. \$\endgroup\$
    – Matt Grum
    Jul 5, 2011 at 9:15
  • \$\begingroup\$ @Matt: 1) I'd say your examples still constitute less than 5% of all pictures most people take, so I think my 95% is a reasonable approximation. 2) At least as your first one would normally be practiced, it doesn't fit with what I was classifying as ETTR anyway (and I doubt I'm unique in that). In a studio you're normally shooting tethered and checking the exposure on the computer not the camera. That changes the situation considerably. Call it ETTR if you like (and I won't argue) -- my argument is primarily against the on-camera histogram, not the idea in general. \$\endgroup\$ Jul 5, 2011 at 10:21
  • \$\begingroup\$ 5%?! I'd say a conservative estimate is that static scenes where you have time to check your exposure after a shot (and shoot again) make up about 50% of shots for a typical amateur photographer, some photographers more like 99%. I'm not saying they do check the exposure after every shot, just that they have the opportunity to. ETTR means increasing the exposure presumably for some gain (e.g. better SNR), I don't see how shooting tethered makes any difference to that definition. \$\endgroup\$
    – Matt Grum
    Jul 5, 2011 at 10:59
  • \$\begingroup\$ I can't imagine where you'd come up with 50% being static scenes. An awful lot of photographs include people, which rules that out. Even without people, clouds keep most scenery from being truly static. Even among pure landscape photographers, it can count as 99% only if you define "static" so loosely that it becomes meaningless, and the "sunny f/16" rule qualifies as "ETTR"! \$\endgroup\$ Jul 5, 2011 at 11:07
0
\$\begingroup\$

Expose To The Right is often used in the wrong setting and is often a redundant exercise. Consider this, you have a scene where 18% exposure on say, the subject's face, is possible at a shutter speed of 1/125th at ISO 200 with an aperture of f/4.0.

You then think, "Oh, I should expose to the right", and consequently dial down up the shutter speed to 1/100th of a second and bring the aperture to f/3.5 to get maybe a stop more light (rough guess). What you've effectively done is use ISO 100 exposure settings while you're still at ISO 200 and now, while your exposure is now a stop brighter than it will be, you're still getting the same signal to noise ratio as any other ISO.

In the end you're going to end up with 99% the same exposure as if you also brought the ISO down to 100 and took the photo that resulted in the normal exposure and didn't have to bring down the exposure in post.

I guess for all intents and purposes, ETTR should really only need to be practiced when sitting on the lowest sensitivity as you're inherently pushing the SNR ratio beyond it's normal limit. Only at the lowest sensitivity will a higher SNR be useful in post.

\$\endgroup\$
4
  • \$\begingroup\$ You ought to get a slight improvement in shadow noise in your example (amplifying the signal before digitization raises any shadow detail above the noise floor), but yes expose to the right is probably redundant if you have plenty of light (which is the case if you're able to shoot at ISO100). However if you're working in low light it can have a very noticeable effect on noise so I completely disagree with your last statement (that ETTR should only be used when you can't reduce ISO). \$\endgroup\$
    – Matt Grum
    Jul 4, 2011 at 10:03
  • \$\begingroup\$ ...cont It seems your argument is that if you are above base sensitivity, say at ISO800 f/4.0 1/100s then if you expose to the right by a stop and shoot f/4.0, 1/50s then you could have just shot at ISO400 f/4.0, 1/50 and obtained the correct 18% metering exposure. So by exposing to the right all you have done is increased ISO for the same amount of incoming light. \$\endgroup\$
    – Matt Grum
    Jul 4, 2011 at 10:17
  • \$\begingroup\$ ...cont However it's possible to demonstrate that by increasing ISO for a fixed shutter/aperture you do improve signal to noise ratio as there's a fixed component to the noise that happens after the analogue amplification step, and by amplifying the signal against a constant noise level you get a bigger signal to noise ratio, clearly). Thus ETTR offers benefits at all sensitivities. \$\endgroup\$
    – Matt Grum
    Jul 4, 2011 at 10:17
  • \$\begingroup\$ Good points. I think in terms of the benefit of analogue vs digital ISO amplification would be the only primary reason to shoot that higher ISO in camera. At least from a logical point of view. \$\endgroup\$ Jul 4, 2011 at 11:35
0
\$\begingroup\$

I almost always use spot metering and "expose to the right" WITHOUT CLIPPING. By doing so, 99 out of 100 I get my exposure "technically correct" Unless I am after some special effects, very little is needed in post editing

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Can you describe a little more how you do that? Without that, this doesn't really answer the question. \$\endgroup\$
    – mattdm
    Aug 27, 2015 at 16:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.