2 Apr
1998
2 Apr
'98
8:41 a.m.
Shawn,
I think your solution is getting a bit too...involved. Yes...overkill.
It is my _understanding_ that the Fotolook "Descreen" function scans the
image at your target resolution but then blurs each channel slightly in
order to eliminate the pixelization (since accurate technical data from
Agfa is *totally* a figment of my rich fantasy life, please realize that I
could be dead wrong on this one)...but I digress.
An easier solution to do your own descreening (sometimes this works better
than the scanner's software), is by scanning at a normal setting (100%
sharpness), at a much higher output resolution (4x)...then blur just enough
to break the pattern...now reduce the file size to your target resolution
and run an unsharp mask. Beautiful.
I ran a test for a lores cover scan:
1) scanned @300dpi, blurred, reduced to 72dpi and unsharp masked @75%
2) scanned @72dpi with a descreen set to 175lpi (we print at 150)
Since Fotolook automatically sharpens to compensate for the descreening
loss, scan 2 was oversharpened and contained some artifacting...scan 1 was
gorgeous.
van
note: Another trick is to actually rotate your image on the scanning bed
(like 20? or so) which usually knocks out the moir?...so long as the
corners don't get clipped. (I think we should return to the regularly
scheduled programming now.)
Well that was my experience with that particular
scanning software but I just
want to know how they did it so I can reproduce the results in something
else.
I was thinking if your CCD has 300 DPI and you scan at say, the dot frequency
of 133 dpi, does it really sample all 300 pixels in an inch and then resample
downwards or does it just sample the CCD pixel that is closest to the dot
being scanned? If it does the former, then scanning at 300 DPI and then
down-converting to 133 DPI in Photoshop should be exactly equivalent; if it
does the latter, then I would expect really bad results because sometimes the
CCD element being sampled falls on a screen dot and sometimes it doesn't. In
reality the dot size is supposed to correspond to the darkness of that color;
so if you had an optical means to sample the entire "cell" consisting of the
screen dot plus the white space around it, the whole cell and nothing but the
cell, then you could get the color right for that cell, but the cells are not
an orthonormal array; the screens are slanted (and to make matters worse,
differently for different colors). So I think the optimal solution might be
to sample at a very very high resolution (many pixels per screen dot), edge
detect the screen dots, use that information to break up the image into cells,
average the darkness for all the dots in each cell, use that color to fill the
entire cell, and then resample that image downwards to the maximum achievable
resolution (whatever that is). As long as screen dots don't overlap you could
even do that separately for each color. I don't know if any existing software
does that. Maybe Fotolook did. Or maybe that's overkill; maybe just a few
samples per screen dot is enough. But I'm pretty sure skipping any of the
CCD's available dots is not a good idea, and I don't know if that's what my
scanner does when I tell it to sample at 133 dpi. (I have an HP IIcx.)
........................................................................
@
/
/ Shift Lever
(D)/
\===================================== @ ================ Floor Plan ===
BMW |- - -Phase Shifter- - - -|--/ Get Wired!
- ------------]=[]@----------------------@ 415.276.4979
Trans- ] ]](A) Toll Free 1.888.208.6655 (B) ? (C) Rear Connection
mission ]]]]]]]]]]]]Driveshaft]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
] ]]
- ------------] web superstation of the stars...
van burnham http://www.futuraworld.com
production manager
wired 520 third street fourth floor san francisco ca 94107 united states
........................................................................
for immediate emergency wireless access send email to van-page(a)wired.com
van(a)wired.com van(a)futuraworld.com pingpong(a)spy.net vanburnham(a)aol.com
Not to ba a prude...but what on earth does this have to do with the
discussion regarding the collecting of classic computers? Archiving?
Not much but it is at least an interesting tangent isn't it?
(Sorry if we're boring anybody.)
--
_______ KB7PWD @ KC7Y.AZ.US.NOAM ecloud(a)bigfoot.com
(_ | |_) Shawn T. Rutledge http://www.bigfoot.com/~ecloud
__) | | \_____________________________________________________________