I guess the title I've chosen for this post could be seen as perhaps a little hyperbolic, but we're doing pretty well - ten to nine in the morning of the first day and there's already something interesting to report.
On the face of it, this would appear to be one of those witchcraft/black-magic moderated techniques which shouldn't actually be possible, but according to a couple of images I've just seen it's now possible to recover sharp images from defocused ones if the camera has received what seems, on the face of it, a fairly trivial modification. Grass Valley's Chief Scientist of Imaging, Peter Centen, described the insertion, at the iris, of an etched metal plate with square holes rather resembling a 2D barcode, a sort of digital pattern referred to as a coded aperture.
The difference made to the image was trivial in the fairly small and indistinct demos provided, although I'm sure it could become visible in bokeh. The clever bit is that it allows some fairly hairy mathematics to be done in order to recover sharp image data. The given demos were pretty convincing, although I suspect not quite perfect given noise and other inevitable degradations that always conspire to make these things less than ideal.
Describing exactly how this is done is beyond the scope of a quick forum post, but the techniques behind this are called deconvolution. People familiar with the full meaning of the phrase "modulation transfer function" will understand what becomes possible when you have an MTF which has zero nodes in it.
As with anything I suspect it'll turn out to be imperfect, and quite dependent on the quality of the original photography, but it may at least be a very good reason to shoot high resolution, 4- or 8K material, even when the distribution will be HD.
PS - We weren't allowed to take photos in the presentation, but it was part of the SMPTE-sponsored digital cinema tech summit, and all the presentations will soon be available at the SMPTE site.