L
lkaven
Guest
Indeed, thanks a lot Luke! When I have some free time to spare I'm definitely going to try executing the technique you've explained.
Just one more question, if you don't mind sharing with us, what exactly does the merging of multiple HDR images achieve better than tone-mapping from a single one - what are the perks?
First to untangle terminology a bit. A single shot exposure is a low dynamic range image (LDR), something that typically has 8-10 stops of usable dynamic range in it. With a good photon counter like the D3, you can get surprisingly clean capture right down to the lowest bit, so there is quite a bit of detail in the shadows of a very good LDR capture. Using tonemapping techniques such as with Photomatix, you can remap those tones, which is often used to bring the shadows up into the midtones and the highlights down to give a compressed tonal range, and sometimes to very good effect. This is especially good for spot work when used judiciously. Just brush in parts of the tonemapped image along with the original. In current terminology, this is called "pseudo-HDR".
An HDR "image" is actually an image file with 32-bit floating point values that scale from pitch blackness to (roughly of course) the surface of the sun, and everything in between. The Radiance format is the one that is most often used, but there is also a floating point TIF representation. This "image" can contain too much range of luminosity to display on ... anything. To make it, you have to collect data, but your camera only collects 8-10 stops at a time. You want to collect maybe 16-20 or so stops of light information, so you have to take 5 or more images two stops apart. Then you take those LDR images and convert them to HDR by feeding them to, say, Photomatix, which is able to read all the exposure information from the EXIF and figure out how to normalize the light values in each exposure, average them together, and come up with a 32 bit floating point number (one each for R, G and B).
Since you can't display this HDR image on anything, you have to come up with a way to map the light values onto visible tones on your display, paper-ink set, or what-have-you. That's the tonemapping step, and there are several ways to do it, the best of which involve computations on local features that sort of mimic what your eye does. Overall, tonemapping is just way to determine how to fit ('map') the huge range of light values you've collected onto the much narrower range of the viewing medium. When you think about it, the problem isn't that different from what painters and illustrators have had to do for centuries, which is partly why HDR images can look reminiscent of paintings and illustrations.
Best advice...always shoot uncompressed RAW files, 14-16 bits if you got 'em. Capture to 16-bit TIFs with the most neutral, linear capture you've got. Don't try to spruce up the images in this step with tone curves, etc. Don't use JPEGs--they have very low dynamic range. And don't bother doing white balance until after tonemapping.
Sooner or later, you'll be able to do this kind of stuff in camera, I think.
Take care, Luke
