Exposure Bracketing via Aperture

The current standard is to vary shutter speed for shooting exposure bracketing sequences. If you'd vary aperture instead, you would also vary the depth-of-field, which is a bad thing because then you get fringing artifacts when you generate the HDRI.

But not anymore. Some crazy folks from the University of Toronto figured it all out, turning this obvious weakness into an advantage! Their new algorithm can not only avoid registration artifacts, they can even use the slight variances in the depth-of-field to create some kind of Über-HDRI where you can set the focus anywhere you want.
Projects website

Their method reads like inverse ray-tracing to me. Apparently, they analyze the tiny variances caused by aperture changes to split the image into several layers. These layers are then treated like cutout-cards in a 3D space: You can set them all in focus, or specify a tight focus range that is on one specific layer, or even somewhere between these layers. Good stuff!
Check out the movies on their project's website, and even if you're not a math fanatic you can safely read the first and the last page of their paper.

Next Page