I’ve previously posted about HDR here on the blog, but it has always regarded taking multiple exposures of the same scene and combining them. This time I have been experimenting with using multiple raw conversions of a high contrast portrait to create an image devoid of any noise artifacts (or maybe devoid of most artifacts). Both of the above images were created from three conversions of the same raw file, using 2 stops of exposure difference between each file. I then converted the resulting hdr images down to 16 bits using Local Adaptation and curves. I would specify the radius and threshold settings, but I have found the radius varies with resolution, and threshold varies with each image (tending to be between .5 and 2).
The lighting for this was a large softbox camera right slightly above the subjects eyelevel, a black nonreflector camera left (as close to the subject as I could get without it in frame), and a kicker light behind the subject camera left feathering onto the background about a foot above the subjects head.
They look good. Is there something to compare it to, like, a single conversion of the raw?
Although I could post a before shot, I don’t think it would really be analogous for comparison. The conversion from 32 bits to 16 bits allows you to make curve adjustments, which effect the contrast of the final image. Perhaps attempting to adjust a non hdr image to match the hdr conversion and then comparing the shadow noise would be the only way to look at it empirically . . . but I am open to suggestions.