Archived posting to the Leica Users Group, 2004/10/27
[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]Hi Frank, just a bit of lateral thinking - what would happen if there was (is there perhaps?) a method of limited randomisation or random shifting of pixels within a defined grid? Or is this what happens with grain emulation in PS?. Would we get a digital result undistinguishable from analogue? I'm thinking on the basis of experience with the many interpolation, noise reduction and signal enhancement processes used in geophysical processing (seismic data). Which, when we get down to it, is only another form of signal processing just like image manipulation. For example we distinguish between random (digital noise/hiss) and ordered (artefacts) interference in seismic data, and all data are in more or less regularly sampled 2 or 3D coordinate systems, surely there must be some applications also applicable to photography. An interesting test would be to take identical images with digital and analogue and subtract one from the other to enable us to see exactly what and where the differences are. Assuming a digital resolution of the same order as the grain density on film it is conceivable that the analogue negative added to the digital positive would cancel out those locations where there is data correspondence (eg positive value + corresponding negative value = zero) and leave us with a "noise image" from which a function for noise reduction or signal enhancement could be extracted with reasonable ease. Either removing the digital "fingerprint" or adding the analogue fingerpint to an image to "analog-ise" any digital data set. Taken further, this could be used to give the film typical appearance and character for any film type to a digital image - Fuji, Kodak whatever. Sorry to waffle on about what's probably of no practical use at all, but who knows......... Douglas Frank Dernie schrieb: > Hi Douglas, > My point was exactly that, one does not produce any new grains on the > negative but unless there is an identical pattern and number of grains > on the print paper, with identical magnification to that chosen in the > enlarger, the print will not be identical to the negative, and > assuming more grains in the print paper than the negative then there > is some similarity with digital upsampling, though it is random rather > than mathematically defined. > Frank > On 27 Oct, 2004, at 00:16, Douglas M. Sharp wrote: > >> >> >> Frank Dernie schrieb: >> >>> Hi Phong, >>> I completely agree. Since film is the most digital of "analogue" >>> media, in that it consists of discrete grains, in a somewhat random >>> pattern, but still basically so many grains per square inch >>> (depending on film). When the film is exposed the grains are >>> effected depending on colour, light intensity etc but there are >>> still a fixed number of grains. When the negative is enlarged a new >>> sheet of "analogue" material is exposed in the enlarger. Unless this >>> sheet of material has identical grain pattern and number of grains >>> as the original negative, and the enlarger lens is perfect, the >>> resulting enlargement has -must have- interpolated information in it >>> and is changed from the original. >>> Frank >>> >> True enough but I think in the case of analogue/film/negative we should >> get away from the term "interpolated".In the technical sense the >> definition >> is that a new, narrower, grid is constructed from a grid of original >> information. In the case of digital information the interstices between >> the existing >> value grid are "filled-out" with new values containing information from >> the surrounding or neighbouring values. >> The degree of influence (weighting) of the neighbouring real data on the >> resulting value is governed by processing algorithms or simple matrix >> filters. The simplest form would be to take an equal proportion of each >> of the four corner values of a grid to produce a reasonable >> approximation to what a real value might be at the centre. More >> complicated methods also take into account the neighbouring groups of >> values, these can be overlapped or weighted (binning) to produce an >> albeit less reliable but more realistic result including trends within >> the data.. >> >> In the case of a negative or film the more or less random distribution >> of grains shows no regular structure, the "data" may "look" different >> after enlargement >> but the finite number of grains remains the same. An interpolation per >> definition at source doesn't take place in so far that we are not >> creating any new grains.. >> Douglas >> >> >> >> _______________________________________________ >> Leica Users Group. >> See http://leica-users.org/mailman/listinfo/lug for more information >> > > _______________________________________________ > Leica Users Group. > See http://leica-users.org/mailman/listinfo/lug for more information > -- Ihre bevorzugten Shops, hilfreiche Einkaufs-Hilfen und gro?artige Geschenk Ideen. Erleben Sie das Vergn?gen online einzukaufen mit Shop@Netscape! http://shopping.netscape.de/shopping/