"Expose to the Right" May Be Wrong
You may have encountered the rule of thumb to "expose to the right," intended to improve the quality of digital photography. You may even know what it means and how to apply it. But is it truly good advice today?
To answer this, it's necessary to cover a little background on how raw digital images are actually captured and recorded. Every digital camera captures raw images, regardless of brand or how much you paid for it. Yours may let you save that raw image, or it may automatically convert it for you to a jpeg image before saving it to your memory card. Either way, it starts out with a digital sensor covered with millions of light sensitive photosites that work pretty much the same way as in every other digital camera. Some sensors are more sensitive than those in other cameras, and some are more accurate, but they all work on the same basic principles.
Each photosite sees a very odd version of the world in front of it, and in only a single color, red, green or blue. Together with data recorded by neighboring photosites that capture light of different primary colors, a composite RGB image gets interpolated. But more important to the topic at hand than the color of light each photosite records, is the intensity of light needed to register increasing brightness.
Human vision sees doubling or halving of light intensity as being equal increments of brightness. On the surface, this might seem somewhat counterintuitive but you probably actually know this fact anyway since it's indeed quite fundamental to how exposure works for photography. Each doubling or halving equates to a "stop" of exposure and impacts how we choose an acceptable aperture, set the shutter speed or pick an ISO setting. Photographic exposure is all about doubling and halving.
But that's not the way a digital camera sensor's photosites see things. Each one merely counts how many photons of light strike it during the time the shutter is open. Double the brightness and you'll get a number twice as large. Double it again and you'll end up with a number four times as large as the original, and so on. This direct, linear relationship between brightness and the resulting value is known as "linear gamma," for those looking to enhance their dinner party conversation vocabulary.
This mismatch between how your camera sensor sees brightness and how you see it has a curious side effect. Suppose your camera records 12-bit raw images. Some current models expand that to 14-bit raw, but the concept here is the same. Regardless, there's a finite range of available values that you can end up with from any single photosite. If you break down that available range by stops, an increasing percentage of the total goes to stops at the upper end of the brightness scale, and a conversely decreasing percentage is available for stops at the lower end. Since each stop corresponds to a halving or doubling of true brightness, fully half of the available values would go to the uppermost stop in a linear gamma system. One half of the remaining (or one quarter of what we started with) would go to the next stop down, and so on. This would seem to be bleak news for the shadows in your image where few discreet values will remain to record data.
It's this side effect that led clever photographers to the idea of exposing to the right. If you're capturing an image that lacks any truly bright areas, you're wasting much of the possible range of possible values and are artificially limiting yourself to not using the full range. If, instead, you somewhat overexpose that image so that the brightest areas result in values within the upper stop of the available range, the entire image will have quite a few more available values in each stop that can be used. Pushing the exposure by one stop boosts the exposure of every stop, giving each one twice the number of available values to record image detail than would otherwise be the case when "correctly" exposed. More available values mean a more accurate image.
Of course such an image would naturally be one stop overexposed since you did so on purpose in order to "expose to the right." To correct the exposure, you'd then need to lower the brightness by one stop later in Photoshop or Lightroom to return the final appearance to the intended correct exposure. Given this, the obvious question is "what's the point?" Overexpose by one stop in-camera, then push it back down one stop on your computer afterwards leaves you right back where you would have been had you exposed it "correctly" in the first place.
The answer to this seemingly pointless over-then-under exposure strategy that is the essence of "expose to the right" is noise, or rather signal to noise. To one degree or another, every digital camera is susceptible to digital noise. With few bits available in the darkest stops, every camera will eventually succumb to noise. When low exposure pushes technology to the limits the result at some point will be a signal with such low information content that it becomes indistinguishable, and thus effectively masked by, noise. Noise is always present, but with enough signal, it become irrelevant. It's like talking louder in a noisy crowd so the person next to you can hear you versus whispering and remaining unheard or poorly understood. In much the same way, the idea behind "expose to the right" is to increase the available signal to noise ratio by upping the strength of the information content above that of the background noise to result in a clearer, cleaner image. When you later push the brightness back down on your computer to return the net exposure to where you want it, you will actually be pushing down the noise level as well, retaining or even enhancing the improvements you worked so hard for in-camera.
All this is entirely valid, and it does work, but it's not without its competing considerations, of which there are basically two.
First off, any time you intentionally overexpose an image you risk burning out the highlights. Just as was true of slide film before it, digital image capture is notoriously unforgiving of too much overexposure. Regardless of the range of brightness values your camera can capture, there are always limits. In basic eight-bit imaging you are limited to values from zero to 255. No matter how much you may want it, there is no 256 or greater. Despite the peculiarities of linear gamma raw capture outlined above, the same thing still holds true. If you envision the brightness scale stretching from pure black on the far left to pure white on the far right, there are impenetrably hard walls on both ends. When you hit the wall on the far right, you simply can't go any further. Pure white is pure white. You can't go any further to the right. And even if you think your image is dark enough overall that you can expose it further to the right to cleverly gain some advantage over any possible noise present, you may have missed some specular highlights or other small bright areas that may be present. Push them even brighter and they flat-line against that hard right side wall. Decrease the exposure later in Photoshop and you'll find the area around those bright spots doesn't respond well. You'll end up with visible hard edged transitions to pure white where there shouldn't be any burned out areas at all or at worst smoother gradients to white. It can ruin an image. This won't impact every image that you expose to the right, but it can be a real bummer for others. Trust me.
Given this, the second competing consideration I wanted to point out is that newer digital cameras are inherently less susceptible to noise in the first place. The march of technology continues to advance after all. I remember my first digital camera, a Nikon Coolpix I bought thinking that it could serve as a convenient supplement to the film cameras I mainly used way back when. I was generally quite disappointed since virtually every image suffered from noise problems in the shadows. It was great for casual shooting when I didn't want to deal with shooting a roll of film and getting it processed, but for anything serious, I was going to stick with film.
But over the years, things of gotten remarkably better. Only when shooting in extremely low lighting conditions is noise an overriding concern these days. Not only has sensor resolution increased with each new model released, so has sensitivity. Greater sensitivity means less worry about noise overwhelming the image signal.
So if there's little need to take extraordinary measures to deal with noise these days, there little incentive to risk possibly ruining an image by needlessly burning out highlights by exposing to the right. Even if that doesn't happen, boosting exposure only later decrease it back to where you started remains extra work with generally no visible payback. The theory of "expose to the right" remains as valid as ever of course. It's just that these days there isn't much need for the extra work or justification for risking unintended consequences.
Unless you've learned what your camera is capable of and what it isn't, and are pushing the bounds of those capabilities knowingly, my advice would be to target your in-camera exposure to what you want the results to be. Although it may once have made sense, "expose to the right" may in fact be wrong today.