Should I shoot underexposed photos?

Question

Someone more experienced than me said that he always shoots so that his photos are underexposed (darker) and I should too. The reasoning being that you can always increase the exposure in postproduction, but you can never get the details back from overexposed parts of the photo.

But if I think about it I think that you lose the dynamic range with this technique. So I have three questions:

1. What should I do? Should I do it that way, or does it depend on the circumstances (like if I have enough time to test for proper exposure, etc.). Or should I rather try to learn to "guess" the right settings (I will fail a lot but also learn a lot)?
2. What do you do? How do you aproach a shoot?
3. How does underexposing affect dynamic range of photos? Does it effectively reduce contrast?

Answer 1

With underexposure, a third of a stop won't hurt you much. Two-thirds is not great, but not that bad. A whole stop means you're doubling your noise. Two stops and you're quadrupling it. So underexposure is by no means "free", but there are grades of it.

With overexposure, any amount will start to clip. Whether that is very detrimental will depend on the subject matter, but it's generally more damaging to your image than the noise of underexposure.

Now to your questions.

• What should I do? Should I do it that way, or does it depend on the circumstances (like if I have enough time to test for proper exposure, etc.). Or should I rather try to learn to "guess" the right settings (I will fail a lot but also learn a lot)?

No, don't underexpose. Aim to expose correctly. If you are not sure if you've exposed correctly, then it's better to underexpose a little than overexpose a little, that is true, though it still won't get results as good as exposing correctly. If you have the opportunity to try again if you make a mistake, it's better to aim for a correct exposure and try again later if you don't get it.

• What do you do? How do you aproach a shoot?

I try to get it spot on. I check the display on the LCD, and might check the RGB histogram, and if I've under- or over- exposed then I take another one. Of course this is not a luxury you have if you're doing photojournalism or anything like that, where you can't re-try the shot.

• How does underexposing affect dynamic range of photos? Does it effectively reduce contrast?

Underexposing reduces dynamic range by removing some range in the darkest areas, such as shadows. So underexposing by one stop reduces your dynamic range by one stop (ie, by half). The noise floor raises, so the loss in dynamic range is expressed as an increase in noise in the darkest parts of the image, obscuring some of the range that otherwise would have shown dark shadow detail.

I wouldn't say that it reduces "contrast" as such. It does reduce the dynamic range - that is, the relative distance between the darkest and lightest detail you can reproduce without it being obscured by noise. But I wouldn't describe that as primarily an effect on contrast - depending how you use the word "contrast".

Answer 2

There's been alot of talk of ETTR which is the opposite of what you're talking, but not much about underexposing. Basically, no, you shouldn't. On most sensors the dark parts are by far the noisest parts of the image, and pushing that in post is just going to make it noisier. You can't recover from pure black either.

The reality is you should expose "properly" - get the creatively correct exposure for the picture you want. Exposure is a creative tool as much as the rest of the tools in the photog's toolbox. With digital cameras, you get instant results, just take the picture how you think you want it and chimp. In critical situations, where you may only get one shot you're going to need to take several pre-exposures of the surrounding area and combine it with experience to get what you want.

I'm not a fan of anybody who says you should always decrease or increase the exposure based on what would properly do it in the first place.

Answer 3

I could turn that same logic on it's head as such: "The reasoning being that you can always decrease the exposure in postproduction, but you can never get the details back from underexposed parts of the photo." The parts that are pure black are just as hard to get detail out of as the parts that are pure white. And in reality, with digital, since sensor noise is "worse" in the dark areas then it's harder to get usable detail out of borderline underexposed area than it is out of borderline overexposed area. (because the slight difference between pixels is proportionally huge compared to the value of the pixel itself.)

There is additionally some arguments that the upper range of the exposure band has more room for detail in it. This is where the "Expose to the Right" or "Shoot to the Right" (ETTR/STTR) mentality comes from that a lot of folks follow. The reasoning behind this is mathematical in nature. Basically it says that if you look at light, it's an exponential growth curve, if you move one stop up, you double the light. But if you look at the digitalized version of that in a file, it's linear. Why does that give you more detail in the right hand side of the spectrum? well, imagine we have 7 stops of light to spread across a 14 bit raw file. In order to do that the 4th stop of light has to have 2x the number of values in it than the 3rd stop of light. Say the 3rd stop of brightness runs from 128 to 256 (that gives it 128 individual brightness levels within the stop), the next stop of light would run from 256 to 512 (giving it 256 individual levels within the stop), the one after that will run from 512 to 1024, then 1024 to 2048... etc. Each stop of light, in order to maintain the logic that a stop is double the light will have a different number of unique levels of brightness within it, those unique levels are what gives you your subtle shadings, more levels = more subtle shades = more detail.

Now, I should say the above is how it works from a gross theoretical standpoint. The camera makers aren't dumb, they know this quickly leads to absurd imbalances in the quality of their images, so they do skew things in their raw files to give more bits to the darker end than the light end... part of the magic of reading a raw file is having the right curves to apply to them to undo that. So it's not quite as bad as those numbers imply if you're shooting raw. BUT the general pattern they show does seem to hold true. This is the origin of the ETTR concept.

To you specific questions:

1. any hard and fast rule that says you should always do something, is probably wrong. You need to always take the current situation into account. (yes, I realize what I just did there. ;)

2. personally, I agree with the ETTR concept, I usually start with my in camera meter set to +2/3 ev and work from there based on what I see in the histogram. If I ever have to pick between slightly over and slightly under due to technical limitations... I always pick raw + slightly over, because I know I can get better results pulling it down in post than I can pushing it up.

3. In my experience it will reduce the level of detail within an area, and it will add noise. The recovery of those problems could well reduce contrast.

Answer 4

I think your advisor might have meant that when the light levels on scene are so different that you have to choose whether to under- or overexpose, underexposing (losing shadow detail) should be preferred to overexposing (losing highlight detail), since missing shadow detail is less noticeable than bright splashes of solid white.

Underexposing might also be a safe bet when your camera does not show histogram for separate color channels, and some color is distinctly dominating. By "exposing to the right", you might accidentally clip an individual color channel (e.g. red) while others still stay below white.

Answer 5

Both under- and over-exposure are exactly as the name says...not correct exposure, so you end up limiting your dynamic range, if overexposed, you have less noise, but you will wash out highlights; if underexposed, you get more detail in shadows but more noise.

Aim to get it as right as possible in the camera, and then post production tweaking will be minimal and you'll have the widest possible dynamic range to play with - always a good thing!

Answer 6

As I learned the definitions, overexposure and underexposure are always errors. There's some correct exposure that is optimal for your artistic goals. If you expose more than that, it's overexposure. If you expose less than that, it's underexposure. But none of these definitions have anything to do with the camera's meter reading!

If you want to avoid clipped highlights (and that's usually a good idea), and you need to expose 2 stops less than the meter reading to do so, then exposing 2 stops below the meter is correct exposure. Exposing at the meter reading would result in overexposure.

So, by these definitions, no you should never underexpose your photos. But you may need to expose less than the meter reading to obtain the correct exposure—especially to avoid clipped highlights, which are difficult to fix after the fact, as others have described in their answers.

Answer 7

would always suggest you stick to your own technique , may be you are setting the wrong exposure settings. may be you hate yourself for once messing up a good photo with a bad setting , but thats how we tend to learn...

i have always prefered to keeping the post processing to the minimum.. why?

because you start to take photos casually as you feel that it can be improved in PP..if you can take a perfect photo with perfect setting in the camera itself rather than adjusting it in the PP , you have greater control and can make further improvement of your best photos.

what would you suggest? taking a bad photo and making it good ? or taking a good photo and make it better??

i like the later part and thats how i follow it :) happy clicking

Answer 8

I'm not certain this is the correct physical explanation but CCDs are sensitive to not only light but temperature so the pixel value they compute actually is light + some measure of heat. The extra value from heat is called dark count. Your camera tries to correct for this by subtracting off to compensate for the dark count but because it is a random factor it can't be done perfectly, and what remains is called dark noise. Since brightening an image is done by scaling pixel values by a factor the percentage error in a pixel value is the real measure of error (when rescaling). Don't know if this dark noise is the main contributor that makes the dark regions of an image look so bad when scaled up, but it is a candidate, because in bright regions if you have plus or minus 5 noise in each pixel that is a small percentage error, but in dark regions, thats a huge percentage error and would scale up to look horrible.