Bottom line: There is no hard and fast rule.
HDR has been a part of photography, in one way or another, for over 100 years – since the 1880s. But in our modern form of photography, it’s been around for about 10 years. There are all kinds of thoughts and theories about HDR. Some hate it, some love it. Those who hate it typically don’t know a lot about it and are opining based on a very limited exposure to it. Those who love it may do so because they like the surreal, grunge look that – still – is popular in some circles. Others, rightly, view it as a tool among the many available to the photographer.
One debate that continues to crop up is whether using fewer exposures spaced further apart is better or whether more exposures spaced closer together is better. The answer? It depends.
When I started doing HDR a number of years ago, I was in the camp that said more exposures spaced closer together was better. My thinking was that more data was better because it gave the HDR software more to work with. I would busily burn up memory card space capturing inordinate numbers of source images 1/3 stop apart thinking I was on my way to HDR glory.
Clearly I was wrong. As I learned more about HDR I modified my approach. Now, shooting Nikon with the 1 stop limitation on the cameras I have, I shoot at 1 stop intervals. With the introduction of mobile apps to control the camera and provide enhanced bracketing I will likely start to modify that approach.
But back to the debate about the number of source images. As I noted above, it depends. What does it depend on? Glad you asked. It depends on the quality of the source images. What do I mean by quality? I mean the information that each contains. If there is proper information in the 1st and 5th images of a 5-shot bracket then you can use 3 images and get the same result as if you used 5. The images below are an example. In the first, I used 7 source images, 1 stop apart. That’s +/-3 stops.
The exact same tonemapping settings in Photomatix 5 were used. What do we see? The two images look basically the same. There are very minor differences that can be adjusted with tonemapping. Now let’s look at a 3-shot merge using the +/-3 and middle exposures. These are just default settings in Photomatix.
Doesn’t look too good does it? There’s all kinds of posterization in the foliage and sky. The metal on the truck fender doesn’t look good. Shadow areas are overly bright. Again, the same tonemapping settings were used. But even playing around with the settings wouldn’t bring this back to something that looked even reasonably decent.
Three stops was too much on either side of the middle exposure. There wasn’t enough information for the HDR software to use in creating the merged image. Why did the 7 shot bracket work? Because it contained the proper information in those intermediate shots and the software discarded the other information it didn’t need.
Is this a universal result? No. It depends on the dynamic range of the scene. If the brightness range of the scene had been wider and the the +/-3 shots had more information instead of being too extreme, then the merged result would have been better. In this case, because the scene had a range that could be captured sufficiently at +/-2 stops, the others above and below that were superfluous.
Bottom line? Same as at the top of the article. There is no hard and fast rule. You need to adjust on a scene by scene basis. One thing is true though: You can’t have too many source images but you can have too few. Capturing more than you need, within reason, is better than not capturing enough. You can always exclude shots if needed (and delete them from the hard drive if you want) but you can’t magically add extra shots in after the fact. Err on the side of conservatism.