Qualitywise, is there any downside to overexposing an image (within the dynamic range of the camera)?

Question

If i underexpose an image and have to crank up the exposure in post, this will also amplify the noise, resulting in a lower quality image. In this case, it would've been better to expose correctly from the start. I was wondering if there is a similar downside to overexposing an image (from a pure quality perspective, leaving aside the additional postediting workload).

Of course, if I overexpose by using a higher ISO and then turn down the exposure later, the added noise from the initial exposure won't magically disappear. Also, if my image is so overexposed that the brightest areas are clipped, those area's wont be fixable either.

But assume I shoot in RAW on a sunny day. Sunny16 says I can use f/16, 1/100sec, ISO 100 for an even exposure. However, I decide to go with a shutter speed of 1/25sec instead, overexposing my image by roughly two stops. Since I'm shooting in RAW, I have a couple of stops of wiggle room in terms of dynamic range, so even with the slight overexposure, no parts of my image should be clipped (for the sake of argument, let's say this is the case, I know that shooting on the edge like that is not a good habit to get into).

In post, I crank down the exposure by two stops. Will the image quality be any worse than that of a photo shot in the same conditions at 1/100sec (same aperture and ISO)? If so, why? Would it be different if I had stopped down the aperture instead (assuming a lens that doesn't have considerable sharpness issues at lower apertures)? Leaving aside the obvious difference in DoF and movement freezing (also shaky hands, lets say I have a tripod and a shutter release cable) that the different aperture/shutter speed will have.

Answer 1

This is known as ETTR which stands for Expose To The Right. As you correctly described, this will improve image quality as long as there is no actual clipping. The name comes from the fact the the histogram will be skewed to the right without actually touching the right edge.

There is one more reason why this is good which you did not mention. Sensors measure light linearly, this means that every stop of exposure has an twice as many values to represent nuances within it. So by increasing exposure, you will use more of the higher precision stops. Here is why:

Let's assume a 12-bit sensor. It reads values as 0-4095. Each spot is twice as bright as the previous one but sensors measure light intensity linearly. So the highest stop uses values 2048-4095. The next lower stop uses values 1024-2047, going down until you get to a point where the signal is drowned by noise which is why not all 12-bit sensors can actually capture a dynamic-range of 12 stops.

The further right you expose, the higher the ratio between signal and noise becomes, so noise is less apparent. The same noise is still there but because the signal is stronger, it has a less impact. Also as you can see you have basically 11-bits to represent nuances the brightest stop and 10-bits for that of the stop before that and so on.

Answer 2

assume I shoot in RAW on a sunny day. Sunny16 says I can use f/16, 1/100sec, ISO 100 for an even exposure.

This is the correct exposure in the sense that middle grey will be at a good level. That's how auto exposure works.

I have a slightly different rule : expose for the most important elements of the scene without compromising shutter speed.

Shutter speed controls motion blur and shake blur - two issues I've found to be a much greater problem in practice than noise ever was. Also be wary of image stabilization, because it allows you to shoot shake-free at very slow shutter speeds, but it's also easy to drift into motion blur territory that way.

However, I decide to go with a shutter speed of 1/25sec instead, overexposing my image by roughly two stops. Since I'm shooting in RAW, I have a couple of stops of wiggle room in terms of dynamic range

I'd suggest you do not have two stops of wiggle room. That's more than I'd generally expect with RAW. Note in particular that even if you get two stops those two stops will not have the same level of color information and you can get false color in highlights when you pull back exposure in post. This would be most typically seen in e.g. pink sky appearing when it was in fact blue. There are clever raw development algorithms that will try and guess the correct color (by extrapolating from the surrounding areas), but they're not perfect and you need to be aware that in trying this you can create false color problems.

If i underexpose an image and have to crank up the exposure in post, this will also amplify the noise, resulting in a lower quality image. In this case, it would've been better to expose correctly from the start.

This is an argument I see a lot by people who teach ETTR without compromising it, but it's not that simple.

I'd suggest the real issue here is obsessing over noise, which is a common trait in many photographers who concentrate on minute detail (pixels) on screen rather than the really import issues (composition, framing, emotional content of shot, tonal balance). I've never had anyone complain about shots because of noise, but even the most uninformed client will spot a poorly composed shot, or one that's got an uninteresting or engaging composition.

As you correctly point out raising ISO can't be used because you get the increased noise anyway doing that.

You can't generally adjust aperture because it screws up depth of field. Sometimes that's not an issue, but it's removing your control from the shot for the sake of chasing the dragon's tail of noise.

That leaves shutter speed. But changing shutter speed leaves you vulnerable to both shake blur and motion blur. In most shooting I do (YMMV) those would be bigger issues than noise.

So generally I tell people to be aware that ETTR is forcing you to compromise your shooting style.

IMO A better solution to the noise problem is :

• Get and learn to use noise reduction software. Note in particular that luminance grain is almost irrelevant (to people viewing images) in practice and chroma noise (color noise) can often be removed without much compromise on sharpness. Also note that selectively applying noise reduction is generally better than the default approach - just applying it the same way everywhere in an image.

• Selective sharpening is also a good way to reduce the image of noise. Sharpening will generally make noise seem worse (at least when you look at pixel level) and by selectively applying sharpening where it's effective and not everywhere you avoid amplifying noise in areas that don't need sharpening. There's no point in sharpening blurred out backgrounds, but people do it and as a result the low level noise in these areas often becomes much more visible (again, at pixel level).

• Stop looking at pixel level. Just break the habit. Very few photos will benefit from pixel level attention to detail, but an astonishing number of photographers worry about little else (and often ignore more significant issues, like composition, framing and lighting). I've never had a client complain about noise in a photo (and in film days it was much worse than in digital).

Answer 3

Decisions regarding exposure must be made on a case-by-case basis depending on what factors are most important and least important to the photographer. Even within the same order of priorities one photographer's desire to expose an image, or an element within an image, to look a certain brightness level may differ from another's.

The main issues with Exposing to the Right (ETTR) are as follows:

• There is a lack of accurate metering information that shows exactly how much headroom a particular camera may have for a specific scene in the raw data compared to the camera's JPEG engine. The result of the jpeg engine is the way most digital camera's metering profiles are written and the jpeg preview image included in a raw file is the source of in-camera histograms generated after exposure. Even with the exact same camera and settings the amount of headroom raw has over jpeg can vary based on the contents of the scene and the color of the light illuminating the scene.
• If there are disparate amounts of red, green, and blue light in a particular scene many cameras' meters can very easily be fooled into blowing out one color channel, even when the camera's meter and a combined brightness histogram both show proper exposure. This can cause color casts in the processed image, sometimes quite dramatically, in the brightest areas that look pure white before exposure is reduced. A common example of this is a mid-day sky that has a pink cast after it has been pushed when exposed and then pulled back in post processing.
• Clipping in only one color channel can also make properly focused images appear to be blurry and out of focus. This can occur even when the camera's meter shows proper exposure! Sometimes the raw data can be very forgiving when the light meter and/or histogram show overexposure. But sometimes even when the light meter and/or histogram show proper exposure one or more channels may already be clipped at some points in the image.

All of these issues can be overcome, but only if the photographer is aware of them and taking them into account when deciding just how far to ETTR.

Newer cameras with RGB+IR light meters are getting much better than the old standard monochrome light meters when metering lighting that doesn't lie along the blue ←→ yellow color temperature axis that runs across the color wheel (i.e LED "club" lights that have a very magenta cast with almost no green in them).

It is up to each individual photographer to decide what is most important to them in an image and then make the appropriate decisions regarding exposure, acceptable motion blur, DoF, noise, etc. based on the priorities each photographer chooses. For some, the obsession is always with absolute lens sharpness and they insist on always using the lens' sharpest f-stop. For others is always about noise in the shadows. For yet others it is about exact color accuracy.

For example, in order to freeze motion under poor lighting conditions sometimes the desire for little or no noise must yield to the desire to capture anything that is not a blurry mess.

Of course, if I overexpose by using a higher ISO and then turn down the exposure later, the added noise from the initial exposure won't magically disappear.

Nothing magically disappears, but exposing to the right and then processing back to the left can reduce the perceived amount of noise in the shadows compared to shooting at a lower exposure.

Don't assume the lowest ISO setting will always yield the least amount of perceived noise. Likewise, don't assume that the same ISO setting always means the same amount of perceived noise, either. Finally, depending on the camera, even when under the same lighting conditions as ISO is increased sequentially from lower to higher settings the amount of noise may not necessarily increase progressively from one setting to the next.