Olympus-OM
[Top] [All Lists]

Re: [OM] Homepage

Subject: Re: [OM] Homepage
From: Kenneth Sloan <sloan@xxxxxxxxxxx>
Date: Tue, 16 Mar 1999 12:59:31 -0600
Finally, a topic on this list that falls within my professional expertise!

> From what I've read, you scan at twice the resolution you want to end up
> with.  When you resize and halve the DPI, pixels are averaged together. 
> Distorted pixels are eliminated in this procedure and moire (?) pattern
> (if any) is often eliminated. Resizing is always down from large to
> small. 
> These are instructions which came with my Visioneer scanner.  I've
> played around with scanning on these terms and the results have been goo
> dso far.  Thanks for your help.
> 
> John

Following the instructions that come with your scanner is good policy -
but it also helps to understand the basic mathematics.  The first rule
of scanning (sampling) is "Low pass filter before you sample".  Almost
all problems disappear if you follow this rule.

Ideally, your scanner should (optically) integrate over a small area
for each digital sample.  If you scan at a high DPI, the small
areas should be very small - if you ask for a lower resolution scan, the
integration should be over a larger area.  If your scanner does not do
this (sorry, I'm not up on the details of current consumer-grade
scanners), then you may be stuck with ONE sampling area size.  In this
case, there is ONE "best" scanning resolution (higher resolution scans
will not really add any information - lower resolution scans may
introduce aliasing).

If your scanner instructions tell you to scan at twice the desired
resolution and then downsample with software, .... I hope you got a
bargain on the scanner.  This smacks of cheap optical design.  A good
scanner should provide a clean digital image without ANY software
manipulation. 

Software which downsamples digital images SHOULD properly filter the
image as it downsamples.  Good software will do a better job than simply
"averaging together" pixels - but that's a reasonable description of
what needs to be done.  Notice that this is equivalent to scanning at
the lower resolution with a bigger "spot size".  It's probably better to
do it optically - but if your hardware doesn't do it, most software
should do a competent job.

GOOD software should also allow you to upsample (to a higher resolution)
without visible artifacts.  However, I haven't found any consumer-grade
commercial software that does this well.  So - the advice to always
re-size from hi-res to lo-res is appropriate.

So..bottom line: I recommend that you (experimentally) determine the
highest scanning resolution that provides useful information (a good
test image will help here).  Then, scan everything at that resolution.
Let the software (filter and) downsample to the resolution appropriate
to the image and your intended use.

[you should *never* see Moire patterns in a properly scanned image - if
 they are not in the original.  If you can see a Moire pattern, it is
 likely that you are scanning at low resolution with a "too small" spot
 size.  This is good evidence that your scanner does NOT adjust the spot
 size as you adjust the resolution - or that you are not tweaking the
 controls properly - RTFM.  If you see the Moire pattern and "eliminate
 it" by averaging and  downsampling, you are alleviating the problem -
 but you are NOT fixing it completely.  ALWAYS scan at high enough
 resolution so that you NEVER see such patterns at ANY resolution.]

-- 
Kenneth Sloan                                            sloan@xxxxxxx
Computer and Information Sciences                       (205) 934-2213
University of Alabama at Birmingham                 FAX (205) 934-5473
Birmingham, AL 35294-1170   http://www.cis.uab.edu/info/faculty/sloan/


< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >


<Prev in Thread] Current Thread [Next in Thread>
Sponsored by Tako
Impressum | Datenschutz