High Dynamic Range (HDR) images are everywhere. You can tell this is true because many are so obvious. But it is also possible to get natural-looking HDR results with a bit of attention to detail.
As background, one of the fundamental limitations of cameras of all stripes is that they can only capture a limited range of brightness in a single shot. It is true that modern digital cameras have improved matters somewhat when compared to the exposure latitude possible with. But when compared with what the human eye is capable of, the problem largely remains largely unsolved.
When we look at a scene, an intricate interplay between eyes and brain cooperate to create what we see. Pupils dilate and contract, and input from other senses merge with signals from optic nerve pathways to create a seamless interpretation of the world around us. "Seeing" is just the name we give to our sensory impressions when eyesight forms a primary component. But a camera doesn't interpret; it merely records. If the aperture is open sufficiently to see into the shadows, it's probably wide enough to burn out the highlights over the same exposure interval. Stop down the lens adequately to prevent burning out the highlights, and you consign the shadows to murky oblivion.
High dynamic range imaging attempts to address this problem with a radical new paradigm that begins by defining an extremely wide-gamut working space. The idea is to accomodate details far beyond what cameras can record, and beyond the limits of what any monitor can display and printer can print. You then combine information from individual camera frames captured at varying exposures into this space while preserving detail from each. Once merged, HDR software allows you to "tone map" the composite to fit better into something that can be displayed and printed. You can't directly see that HDR working space, but it's there behind the scenes, providing a unifying framework to allow you to combine source images with fidelity in each.
In theory, this sounds great. HDR uses software to create a mathematical model for visual images that isn't limited by the mechanical and technical capabilities of photographic equipment. But it isn't easy to get the source images to add, nor is it easy map the resulting data for output. Both input and output can be problematic when attempting to do HDR.
On the input side of the equation, it's crucial to capture usable data across the entire frame and the range of exposures present. If you fail to collect all the needed data, you'll have a hard time once you get everything onto your computer. All HDR programs can do is to use the best frame possible, even if it may be horribly under or over-exposed. As a starter, shoot a frame that records details in the shadows. Then fire off a shot every full stop from there on up until you're past where the highlights fall. Don't try to cut corners with the highlights. If you misjudge where the brightest portion is in a scene, you're better off overshooting your mark for the final shot than risk finding an even brighter spot later. If you end up finding darker areas than you noticed in the field, they'll be shadows in the result. If you find highlights you missed, you'll be left with burned-out highlights that no HDR can save.
Time permitting, I also try to shoot extra frames, targeting exposure for important areas. That way, I minimize the chances of flaws from object due to movement between frames. On particularly problematic images, I have resorted to manually blending from control shots, replacing the HDR result entirely in certain areas. It never hurts to have a single-shot version of the foreground for this purpose, just in case.
And as a reminder, you don't want to cause such movement yourself. Use a sturdy tripod and a reliable shutter remote. If you don't find it too awkward to tether to a computer or tablet in the field, do so. Depending you're your favorite operating system and camera maker, you should find several software choices to simplify the whole process of shooting for HDR, including adjusting the shutter speed between frames.
On the output side, my main advice would be to use restraint. HDR applications have to use a sleight of hand not dissimilar from sharpening to pull the pieces together. In sharpening, the software selectively lightens or darkens pixels surrounding a perceived edge to make it stand out more clearly. In HDR "tone mapping," pixels across a wider area are selectively adjusted to bring them more in line with the exposure of the next frame in the series. An area that is notably brighter on one end than it's surroundings may get subtly darkened as it nears the other to blend in. It's all quite magical when you start to examine it in detail. Both sharpening and HRD blending are software optical illusions in action.
But there's a fine line between a tasteful magic act and an obvious hoax. Push either sharpening or HDR tone mapping far enough, and you will end up with a garish cartoon of an image. Early consumer HDR software was so crude that it was challenging to get believable output. As a result, more than one online forum sprung up precisely to show off the new "HDR look." Users competed to create the most highly saturated, psychedelic renditions possible. Thankfully, as the tools have improved, most users have shifted back using HDR to creating photographs rather than pop art. Yet it's still easy to overdo it without realizing you've crossed an upspoken line before it's too late. Just as with sharpening, the temptation to overdo HDR is hard to resist. If a little bit looks good, cranking it up a bit more might look better. If you keep at it, you can convince yourself that each ratcheting up of controls is an improvement. Even if you later realize you've been led astray and can moderate things a bit, you'll have wasted a lot of time.
If you're shooting for architecture, you might want to ensure you can see detail throughout the frame. But if you're shooting landscape, you'd be well-served to retain at least some blackness. Shadows are to be expected in the outdoors, both twilight and mid-day. Remove them all through extreme tone mapping, and you'll end up with a flat, lifeless image.
I should also point out that no program can do a good job on every image. As frustrating as it is, it's best to play around with several until you find one you like, but to keep the others in reserve for when your chosen program can seem to work its usual magic with a given image. Sometimes, it can take a bit of tinkering and experimentation to get the best results.
And if all this sounds too complicated, you may be right. As I mentioned at the outset, modern digital cameras continue to improve the ability to cover extreme dynamic range scenes without loss. I've surprised myself more than once by shooting a series of shots for HDR, only to find that a single frame in the middle can work out just fine, all on its own. With a bit of dodging and burning, it's remarkable what new cameras can pull off. With selective editing, detail can be pulled from shadows. Highlight recovery during raw conversion can save otherwise burned-out areas so long as at least one channel retains detail. There are limits, though. When in doubt, shoot for HDR. If you don't need all the extra frames, so be it. But if you shot only one frame, you're stuck with it.