> -----Original Message----- > From: Rob Studdert [mailto:[EMAIL PROTECTED] > Sent: Tuesday, 26 July 2005 11:49 PM > > I believe that his binary grain clump theory is somewhat flawed rendering the > remainder of the argument moot. >
I think the same, and wonder what Reichmann has been smoking or drinking. I've never before heard anyone claim that film grain was binary. Only lithographic film has anything approaching binary characteristics. General purpose films have silver halide particles that grow larger as the exposure &/or development increases. IOW they express their 'analogueness' partly through a gradation in size, and also because a film emulsion layer is usually exposed and developed from the surface down. A slight exp & dev will only produce small grains at the surface of the emulsion layer, while heavier exp & dev will form large grains that fill the emulsion layer from top to bottom. The full tonal range is expressed proportionally between these extremes. IOW: binary film is a bullshit argument. Furthermore, a pixel on anything other than a monochrome or a Foveon sensor is by itself not very informative. To get full colour information a 3 x 3 cluster of pixels is required, i.e. 9 pixels, not 1. It's not at all that simple either, because the clusters are not indivisible units. To get effective interpolation from the gridded RAW colour scheme to a useable RGB file would require the algorithms to be applied in a partial leapfrogging manner, stepped and repeated, back and forth, up and down across the entire image. Not just chrominance but also luminance are affected by this routine (they are not separate channels until Photoshop makes it so). If you shoot for B&W on a colour sensor, you will also be bound by those limitations even though you later reduce the image to greyscale. My arithmetic is provably terrible, so I'll leave conclusions up to you the reader, but the pixel dimensions that Reichmann quotes obviously need to be adjusted to cover a cluster and not just a single pixel, or even a pair of adjacent pixels. The layout of a Bayer grid reveals that the green channel has best most uniform resolution. Each green pixel always has its closest green neighbour touching it diagonally, and is separated by only 1 intervening dissimilar pixel on the horizontal and vertical axes. Red and blue channels resolution is much lower. Their closest same coloured neighbour horizontally and vertically is separated by 3 intervening dissimilar pixels. Diagonally, the gap is 1 intervening dissimilar pixel on one axis, and no intervening dissimilar pixels on the other. Is it any wonder that we get colour fringing. (Incidentally, Fuji's Super-CCD is meant to overcome this problem by rotating the sensor to make red and blue resolutions their best on the vertical and horizontal axes rather than the diagonal axes. The interpolation to a larger file was just a hedge against dodgy algorithms, it has been suggested. New S-CCD Fujis have improved interpolation algorithms and the file size inflation is no longer considered necessary.) Despite Reichmann's ignorant ramble it's obvious that digital capture compares favourably with film, often better but sometimes not. Even in the digicam world where I dwell the results are so good that I haven't loaded a film for more than a year. It's amazing to think that a format half the size of Minox can hold its own against 35mm in many situations. regards, Anthony Farr
