[Tig] A film lifeline?
adrian.thomas at unit.tv
Mon Aug 18 08:47:49 BST 2014
Further to this, the best colour I've seen from a digital camera came from my old Olympus E-1 which used a Bayer pattern KAF-5101C sensor. A Kodak sensor.
Surely we haven't forgotten Cineon et al already? We still use DPX every day. It was Kodak that defined much of what's possible in digital imaging.
50 Gt Marlborough St
T 020 74945700
On 17 Aug 2014, at 12:36, Richard Kirk via Tig <tig at colorist.org> wrote:
> Some people have wondered how Kodak did not see this coming. They did: I remember reading a paper about 1995 where Kodak people said that film had about 15 years left, and sales would plummet in about 2010 (pretty spot-on). They tried for years, and for some while they were the source of the largest sensors (with was 1K square back then). This was fine for satellites, and for the earliest HD digital cameras for people who could afford it, but it was not cheap enough. As Bob K pointed out, they may look like an image engineering clean-room technology company, but what they did best was chemistry; and it is a small step from the flow chemistry of film to brewing beer and wine. Far fetched? I have known of a couple of people who retired from film processing, and now use their experience to run a vineyard. Asahi started off doping something quite different. Anyway, Kodak tried valiantly to make electronic sensors, but in the end old dogs and new tricks did not mix.
> We now have electronic cameras that are probably more sensitive than film. We are close to getting one electron per absorbed photon. Doped silicon is not black, so we could probably increase that by a stop or so. If we could stick the RGB detectors on top of one another like the Foveon sensor, then we might get another two stops because we are using one pixel instead of four, plus perhaps a bit extra for the dead region between the pixels. Together we perhaps have four stops to go. The rest is going to be reducing the background noise, which is probably (a) cooling the sensor (horrible) or (b) using more cunning image processing to clean up the image.
> My best guess is in the short term we shall stick with the Bayer matrix unless someone comes up with a commercial way of getting RGB out of one pixel. People will continue to tweak the sensor, and go to larger areas; but the large visual gains may come from cameras absorbing GPU processors and doing intelligent demosaic and denoise processing that is optimised for their particular sensor. Well, that's just my guess.
More information about the Tig