The point of my original post was to say that while HDR is a useful built-in option, it still requires some basic understanding to use effectively. I want to follow that up by taking a look at one way to improve your in camera HDR photo taking: Exposure.
Exposure is key to HDR photography. Multiple shots are taken at different exposures and then merged in a way that results in an image with more detail than the individual pictures have. When done manually, one might target the exposure for each image to capture detail in a specific area. For instance, with one image you might underexpose the image to pull out the details in what would normally be an overexposed sky and with another you might overexpose to pull out detail in a rock formation seen in deep shadow. Unfortunately, because the iPhone completely automates the over/underexposed captures, all you have control over is the "normal/middle" exposure.
The question I had is whether or not the iPhone's autoexposure was the best starting point for a good HDR result.
The Exposure Test
Photo 3 is pretty close to where the iPhone set the exposure and the HDR result is not horrible. As I mentioned in my previous article, iOS HDR works well to correct some common exposure problems. But can we do better?
The exposure point for Photo 1 was set to the brightest area. This, not surprisingly, underexposes the rest of the frame. The iPhone does a decent job at pulling back some detail.
For Photo 2, I set the exposure point to a darker area of the sky. The result is a dark original and pretty good detail in the HDR. The foreground might be a little darker than I would like, but there is detail there that you don't get in any of the other HDR images.
Photo 4 was exposed based on the extreme foreground area. The sky in the original is completely washed out. Again, very much like Photo 1 (except at the other end of the light range), iOS HDR pulled back some of the detail in the sky, but could not get it all.
Photos 2 and 3 are obviously the best of the bunch. Just eyeballing, my preference was leaning more towards the manually exposed Photo 2 versus the auto exposed Photo 3. This was due in part to the amount of perceived detail as well as the white balance.
While this was all somewhat interesting, I was also curious about how much clipping was occurring at both the low and high ends. Keep in mind that the iPhone spits out an 8-bit JPG so you are not getting the full range of values. Clipped pixels represent image data that is gone forever.
The iPhone seems to do a pretty good job at avoiding clipping at the low end. Surprisingly so (at least to me).
In Photo 1, I expected deeper blacks considering the original was so underexposed to start with. I suspect that the iPhone did not use it at all.
Photo 4 obviously has heavy clipping at the high end.
I was a little surprised at the difference between Photos 2 and 3. I actually expected more high-end clipping in Photo 2. So I am still partial to Photo 2. Opinions will vary for numerous reasons and I could be convinced otherwise with a good argument. Feel free to leave comments with your thoughts.
It appears that auto exposed images can result in a reasonable final product. Not surprising given Apple's "it just works" reputation. However, I have seen nothing to change my opinion that understanding how Apple has implemented their in camera HDR is important to getting the best results possible.
The fact that low end clipping is minimal in all of these leads me to think that I might start edging more towards underexposing to make sure I get as much detail in the high end as possible.
On a whim, I grabbed all of the original images and had Photomatix handle the HDR. I did very little tweaking other than adjusting for some ghosting due to some cloud movement. It's an interesting comparison.
I did a little more tweaking on the Photomatix processed version. Check out the results.