Today I sat down and went about creating HDR images out of the multiple exposures I took on my last photo excursion. It was pretty tricky getting the settings right in Photoshop and it took a couple of hours of experimenting before I felt the output was what I wanted.
(If you just want the photos, scroll down.)
The main problem seems to be that when you press a large tonal range into 8 (24) bits of colour you easily end up with a very dull looking photo.
The fix for this dullness is using curves and
local adaptation when converting from 32-bit to 8-bit. But, with endless possible combinations of settings it takes a while to find something that looks good and realistic.
I finally ended up with these settings:
- Radius: 120 px
- Treshold: 1.0
I also adjust the curve to only just clip the information in the 32-bit image:
Those settings give me this:
And for comparison, the original (metered exposure):
I have put the best shots in a Flickr set.
Getting the shots right
While making the HDR images I realized that I didn’t take dark enough exposures to retain all the detail in the brightest highlights. I used the cameras auto-bracketing feature and did -2, 0 and +2 exposures, except for the last shot I took where I manually did -4 EV to +3 EV at 1 step intervals.
And it is only in that last shot that I can get all the detail around the sun.
So if you’re including the sun or a bright sky, you should probably go at least four stops below the metered exposure, and four above. If you follow Denny Tangs advice here that means that a series of -4, -2, 0, +2 and +4 should do the trick if you’re using an SLR.
If you use Photoshop for merging into HDR I’d love to hear what settings you tend to use.
Edit: Since posting I have decided to also add a very slight S-curve for slightly more contrast. The photos on Flickr have been updated.