On Sat, Feb 16, 2002 at 12:54:41PM -0000, Frits J. W?thrich wrote:
> 100 lines per millimetre is 25.4 x 100 is 2540 lines per inch. Is it OK to
> say that this could be translated to 2540 ppi, or do I miss something here?
100 lines per millimetre refers too 100 lines high density *and* the gaps
of low density inbetween. So you need to sample 200 lines per millimetre
to capture both the lines and the gaps inbetween.
This is not enough. Imagine the image you are trying to scan is offset
exactly half a line with the CCD, so that each of the scanner pixels sees
exacyly half a black and half a white line. You would scan a sheet of
perfect gray. All the detail would be lost.
So you need to sample at at least 400 lines per millimetre to get all the
detail. This is a well-known aspect of sampling theory - you need to
sample at twice the frequency of the signal you want to capture. This is
why audio CD's sample sound at 44KHz - the highest pitch the human ear can
hear is about 22KHz.
That means we need to scan at 10160 ppi to capture all the detail from an
ideal fine-grained negative that resolves 100 lines per mm.
Still this is not the end. This far we assumed that film grain is regular
and rectangular, like pixels. It isn't. Film grains come in various shapes
and sizes. And they are laid out in a random pattern, not a regular grid.
Whether this means that we have to scan at an even higher resolution than
10160 ppi I do not know. I do know that the mismatch between a rectangular
grid of even-sized CCD elements and a random array of variable-sized film
grains can yield a nasty phenomenon called "grain aliasing" at around 2900
dpi, and that scanning at a higher resolution is the only cure.
--
,_
/_) /| /
/ i e t e r / |/ a g e l
-
This message is from the Pentax-Discuss Mail List. To unsubscribe,
go to http://www.pdml.net and follow the directions. Don't forget to
visit the Pentax Users' Gallery at http://pug.komkon.org .
