Next message: { brad brace }: "Re: Resolutions"
Message-Id: <200106041718.KAA25898@dns.ccit.arizona.edu>
Date: Mon, 4 Jun 2001 10:17:24 -0700
From: Guenter Waibel <mailto:guenter@UCLINK4.BERKELEY.EDU>
Subject: Re: Resolutions
To: mailto:IMAGELIB@LISTSERV.ARIZONA.EDU
<pre>
--============_-1220447449==_ma============
Content-Type: text/plain; charset="us-ascii" ; format="flowed"
Hi David & everybody,
that's a really interesting question that's been popping up here and
there in my work: how do you really meaningfully quantify the size of
a file and set an in-house production standard for it?
If you specify ppi, then you might wind up with a lot less data than
you actually want for objects with small physical dimensions. For
example, 600ppi used to be (and still is) a very popular
rule-of-thumb specification for archival quality imaging. However, a
slide (35mm x 24 mm) scanned at 600ppi results in a file of about 1.4
MB, and image dimensions of 827 x 568 pixels - probably not a scan
most institutions would consider archival quality. The 600ppi
standards seems to have been conceived around 4x5 transparencies - it
will yield a filesize of about 21 MB, and image dimensions of 2400 x
3000 pixels. These pixel file dimensions are consistent with the
minimum pixel dimensions that started popping up once projects
started specifying their standards by "resolution in the longest
dimension" - usually, that figure ranges right around 3000 pixels.
Defined by longest dimension, capture of a small object yields a very
similar amount of data as capture of a very large object, deviating
only by the variance in height-width proportion, not by the
difference in absolute size. The file size for the slide and the
transparency captured at 3000 pixels longest dimension would amount
to about 18 MB and 21MB respectively. These are much more consistent
pools of data (compare to 1.4 MB and 21 MB for 600ppi) for objects of
varying sizes.
So the lesson to be learned from all this is that in order to
describe the amount of data captured you always have to put ppi or
pixel dimensions into relation to the original object's size, and
somehow, that process eludes an easy formula.
Both prevalent ways of describing the size of a file (or its
resolution if you want to) have serious conceptual flaws: ppi will
produce substandard captures for small objects and unmanageable
(uncapturable) MB filesizes for large objects. Just try to even think
of a way how you could capture an epic painting covering the whole
wall of a gallery wing at 600ppi :-)
Longest dimension will produce a fairly statis filesize in terms of
MB ( the only variance, as I pointed out, being the height-width
proportion), but all your resulting files will have a different ppi
resolution of the source. While you are capturing 600 pixels for
every inch of the 4x5 in our example of 3000pixel longest dim, you
would only be capturing 300 pixels for every inch of an item the size
of 8x10, etc, until you really wind up not "resolving" your object in
a satisfactory manner at all anymore once they get bigger.
A working solution may be to specify both a minimum pixel dimension
for the longest dimension of the source, and a minimum resolution for
your capture, and then you'd try to use the standard that gives you
the bigger MB filesize. However, at the extreme ends of the spectrum
in terms of source object size, you'll always end up throwing your
standards to the wind and plain old do the best you can do with the
hardware at hand.
I've copied and pasted below a section from the California Digital
Library's Digital Imaging Collections Standards (for download from
http://www.ucop.edu/irc/cdl/tasw/Current/current.html ), which makes
the point for the mixed longest dim / resolution poing I've been
doing my best to elucidate. It also throws a further complication
into the mix, which is the specs for capturing surrogates - in that
case, you have to consider two additional points: (1) the maximum
resolution of the surrogate film stock (see the discussion we've been
having on this list lately), and (2) the original size of the object,
which is NOT the size of the surrogate.
_______
"The intent of the following table is to offer guidelines for
scanning various types and sizes of original documents, so that the
digital master files as captured will record all of the significant
visual features in the original item. Capture resolutions in the
table are based upon the assumption that a scanning resolution of 600
ppi will be sufficient to meet this requirement for most originals in
most collections, apart from negatives and transparencies. Digital
master files which fail to capture some of the visual information
present in the original will presumably become obsolete as image
capture techniques improve over time.
The reflective formats, represented by the first three rows, are
based on 8.5" x 11" originals scanned at 600ppi. The 35mm format has
a resolution standard of 4200 pixels in the longest dimension, as
this is about as much data that most 35mm films can capture. Scanning
the 35mm format, which is 1.5" on the longest side, at 2800ppi will
result in compliance with the 4200 pixel standard. Note, if you
plan to create a film intermediary of the object and then digitize
the 35mm intermediate, remember to consider the size of the original.
Filming an original larger that 5" x 7" with the 35mm process will
not capture all the original's detail. For example, 4200 pixels
spread along the 7" inch side yields 600ppi (4200 pixels / 7"). If
the original was 12" long, the image would be only 350ppi (4200
pixels / 12"), which is not archival quality.
Other transmissive formats, besides 35mm, are represented by the last
two rows of the table. Here the standard is 6000 pixels on the
longest side, based on a 8.5" x 11" original, which yields an image
just under 600ppi image (6000 pixels / 11"). Note again, if you plan
to create a film intermediary and then digitize, you must consider
the size of the original. For example, creating a 4" x 5" negative
at 1200dpi from a 10" long original original yields the 6000ppi
standard, or a 600dpi image (6000 pixels / 10"). Creating an image
6000 pixels on the longest side for a 12" long original would be
digitizing at 500ppi and therefore, would lose detail from the
original image.
Oversize originals such as posters and maps can be especially
difficult to scan at the recommended resolution of 600 ppi. Few
libraries own flatbed scanners capable of scanning originals larger
then 11" x 17," and even if they do, the problems of handling image
files larger than about 120MB are daunting. These problems may lead
to the use of a lower standard of capture resolution, such as 300 ppi
or 3000 pixels in the longest dimension (the "alternative minimum"),
with the understanding that the useful life of the files may be
limited and digital image capture for these objects will need to be
repeated in the future."
______
I hope at least some of what I've had to say makes sense - it's a
tough topic to think through with some stringency. I'd love to hear
what other folks have to say about this issue.
Guenter
>Interesting discussion on resolutions.
>Which raises a question I'd like to throw to the 'thread'...
>
>We are investigating a large scale digitisation project of glass
>plate negatives (GPNs) of all sizes from 1/4 plates to 10"x12", a
>total of 140,000.
>The project is to span over a period of more than 5yrs.
>
>I am researching info for the specification of resolution and quality
>(currently we have been scanning GPNs for 6 yrs and we initially put
>a stake in the ground of 8 megabytes per image - that was a
>respectable file size in 1995!)
>
>At the moment I am considering that a scanned image would have a
>mnimum of 5,000 pixels on its longest edge, this equates at about
>20Meg for any given plate size
>or....
>There is the philosophy that no matter what the GPN size, it should
>be scanned at the same resolution, for example 800ppi.
>(This stands to reason, a 8x10 neg has a lot more information than a
>1/4 plate)
>But of course a 10"x 12" GPN at 800ppi is very large file size -
>100Meg+ at a guess.
>
>The question is
>What approach are other projects selecting?
>
>
>PS
>One point that no one has mentioned
>an average quality scanner used at its highest resolution is not
>necessarily going to produce a better quality scan than that of a
>very good quality scanner at a lower resolution - i.e. I'd rather
>produce a 12 Meg scan that is sharp and has a good dynamic range
>than say a 40 Meg soft scan with clipped shadows and highlights.
>
>
>
>
>
>
>
>
>
>
>
>David Adams
>Team Leader Copying Services
>National Library of New Zealand
>+64 4 4743151
>Visit "Timeframes" New Zealands leading source of heritage images
>http://timeframes.natlib.govt.nz/
>www.natlib.govt.nz
--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Guenter Waibel
Berkeley Art Museum & Pacific Film Archive
Digital Media Developer http://www.bampfa.berkeley.edu/
Digital Imaging SIG Chair, MCN http://www.mcn.edu/visig_subscribe.taf
mailto:guenter@uclink4.berkeley.edu
Phone 510-643-8655
Fax 510-642-4889
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
--============_-1220447449==_ma============
Content-Type: text/html; charset="us-ascii"
<!doctype html public "-//W3C//DTD W3 HTML//EN">
<html><head><style type="text/css"><!--
blockquote, dl, ul, ol, li { margin-top: 0 ; margin-bottom: 0 }
--></style><title>Re: Resolutions</title></head><body>
<div>Hi David & everybody,</div>
<div><br></div>
<div>that's a really interesting question that's been popping up here
and there in my work: how do you really meaningfully quantify the size
of a file and set an in-house production standard for it?</div>
<div><br></div>
<div>If you specify ppi, then you might wind up with a lot less data
than you actually want for objects with small physical dimensions. For
example, 600ppi used to be (and still is) a very popular rule-of-thumb
specification for archival quality imaging. However, a slide (35mm x
24 mm) scanned at 600ppi results in a file of about 1.4 MB, and image
dimensions of 827 x 568 pixels - probably not a scan most institutions
would consider archival quality. The 600ppi standards seems to have
been conceived around 4x5 transparencies - it will yield a filesize of
about 21 MB, and image dimensions of 2400 x 3000 pixels. These pixel
file dimensions are consistent with the minimum pixel dimensions that
started popping up once projects started specifying their standards by
"resolution in the longest dimension" - usually, that figure
ranges right around 3000 pixels.</div>
<div><br></div>
<div>Defined by longest dimension, capture of a small object yields a
very similar amount of data as capture of a very large object,
deviating only by the variance in height-width proportion, not by the
difference in absolute size. The file size for the slide and the
transparency captured at 3000 pixels longest dimension would amount to
about 18 MB and 21MB respectively. These are much more consistent
pools of data (compare to 1.4 MB and 21 MB for 600ppi) for objects of
varying sizes.</div>
<div><br></div>
<div>So the lesson to be learned from all this is that in order to
describe the amount of data captured you always have to put ppi or
pixel dimensions into relation to the original object's size, and
somehow, that process eludes an easy formula.</div>
<div><br></div>
<div>Both prevalent ways of describing the size of a file (or its
resolution if you want to) have serious conceptual flaws: ppi will
produce substandard captures for small objects and unmanageable
(uncapturable) MB filesizes for large objects. Just try to even think
of a way how you could capture an epic painting covering the whole
wall of a gallery wing at 600ppi :-)</div>
<div><br></div>
<div>Longest dimension will produce a fairly statis filesize in terms
of MB ( the only variance, as I pointed out, being the height-width
proportion), but all your resulting files will have a different ppi
resolution of the source. While you are capturing 600 pixels for every
inch of the 4x5 in our example of 3000pixel longest dim, you would
only be capturing 300 pixels for every inch of an item the size of
8x10, etc, until you really wind up not "resolving" your
object in a satisfactory manner at all anymore once they get
bigger.</div>
<div><br></div>
<div>A working solution may be to specify both a minimum pixel
dimension for the longest dimension of the source, and a minimum
resolution for your capture, and then you'd try to use the standard
that gives you the bigger MB filesize. However, at the extreme ends of
the spectrum in terms of source object size, you'll always end up
throwing your standards to the wind and plain old do the best you can
do with the hardware at hand.</div>
<div><br></div>
<div>I've copied and pasted below a section from the California
Digital Library's Digital Imaging Collections Standards (for download
from http://www.ucop.edu/irc/cdl/tasw/Current/current.html ), which
makes the point for the mixed longest dim / resolution poing I've been
doing my best to elucidate. It also throws a further complication into
the mix, which is the specs for capturing surrogates - in that case,
you have to consider two additional points: (1) the maximum resolution
of the surrogate film stock (see the discussion we've been having on
this list lately), and (2) the original size of the object, which is
NOT the size of the surrogate.</div>
<div><br></div>
<div>_______</div>
<div><br></div>
<div><font face="Helvetica" size="-3" color="#000000">"The intent
of the following table is to offer guidelines for scanning various
types and sizes of original documents, so that the digital master
files as captured will record all of the significant visual features
in the original item. Capture resolutions in the table are based upon
the assumption that a scanning resolution of 600 ppi will be
sufficient to meet this requirement for most originals in most
collections, apart from negatives and transparencies. Digital master
files which fail to capture some of the visual information present in
the original will presumably become obsolete as image capture
techniques improve over time.</font></div>
<div><font face="Helvetica" size="-3" color="#000000">The reflective
formats, represented by the first three rows, are based on 8.5" x
11" originals scanned at 600ppi. The 35mm format has a
resolution standard of 4200 pixels in the longest dimension, as this
is about as much data that most 35mm films can capture. Scanning the
35mm format, which is 1.5" on the longest side, at 2800ppi will
result in compliance with the 4200 pixel standard. Note,
if you plan to create a film intermediary of the object and then
digitize the 35mm intermediate, remember to consider the size of the
original. Filming an original larger that 5" x 7" with
the 35mm process will not capture all the original's detail. For
example, 4200 pixels spread along the 7" inch side yields 600ppi
(4200 pixels / 7"). If the original was 12" long, the
image would be only 350ppi (4200 pixels / 12"), which is not
archival quality.<br>
Other transmissive formats, besides 35mm, are represented by the last
two rows of the table. Here the standard is 6000 pixels on the
longest side, based on a 8.5" x 11" original, which yields
an image just under 600ppi image (6000 pixels / 11"). Note
again, if you plan to create a film intermediary and then
digitize, you must consider the size of the original. For
example, creating a 4" x 5" negative at 1200dpi from a 10"
long original original yields the 6000ppi standard, or a 600dpi image
(6000 pixels / 10"). Creating an image 6000 pixels on the longest
side for a 12" long original would be digitizing at 500ppi and
therefore, would lose detail from the original image.</font></div>
<div><font face="Helvetica" size="-3" color="#000000"> Oversize
originals such as posters and maps can be especially difficult to scan
at the recommended resolution of 600 ppi. Few libraries own flatbed
scanners capable of scanning originals larger then 11" x 17,"
and even if they do, the problems of handling image files larger than
about 120MB are daunting. These problems may lead to the use of a
lower standard of capture resolution, such as 300 ppi or 3000 pixels
in the longest dimension (the "alternative minimum"), with
the understanding that the useful life of the files may be
limited and digital image capture for these objects will need to
be repeated in the future."</font></div>
<div><font face="Helvetica" size="-3"
color="#000000">______</font></div>
<div><br></div>
<div>I hope at least some of what I've had to say makes sense -
it's a tough topic to think through with some stringency. I'd love to
hear what other folks have to say about this issue.</div>
<div><br></div>
<div>Guenter</div>
<div><br></div>
<div><br></div>
<blockquote type="cite" cite>Interesting discussion on
resolutions.<br>
Which raises a question I'd like to throw to the 'thread'...<br>
<br>
We are investigating a large scale digitisation project of glass plate
negatives (GPNs) of all sizes from 1/4 plates to 10"x12", a
total of 140,000.<br>
The project is to span over a period of more than 5yrs.<br>
<br>
I am researching info for the specification of resolution and
quality<br>
(currently we have been scanning GPNs for 6 yrs and we initially put a
stake in the ground of 8 megabytes per image - that was a respectable
file size in 1995!)<br>
</blockquote>
<blockquote type="cite" cite>At the moment I am considering that a
scanned image would have a mnimum of 5,000 pixels on its longest edge,
this equates at about 20Meg for any given plate size<br>
or....<br>
There is the philosophy that no matter what the GPN size, it should be
scanned at the same resolution, for example 800ppi.<br>
(This stands to reason, a 8x10 neg has a lot more information than a
1/4 plate)<br>
But of course a 10"x 12" GPN at 800ppi is very large file
size - 100Meg+ at a guess.<br>
<br>
The question is<br>
What approach are other projects selecting?<br>
<br>
<br>
PS<br>
One point that no one has mentioned<br>
an average quality scanner used at its highest resolution is not
necessarily going to produce a better quality scan than that of a very
good quality scanner at a lower resolution - i.e. I'd rather
produce a 12 Meg scan that is sharp and has a good dynamic range than
say a 40 Meg soft scan with clipped shadows and highlights.<br>
<br>
<br>
<br>
<br>
<br>
<br>
<br>
<br>
<br>
<br>
<br>
David Adams<br>
Team Leader Copying Services<br>
National Library of New Zealand<br>
+64 4 4743151<br>
Visit "Timeframes" New Zealands leading source of heritage
images</blockquote>
<blockquote type="cite"
cite>http://timeframes.natlib.govt.nz/>
<blockquote type="cite" cite>www.natlib.govt.nz</blockquote>
<div><br></div>
<div>-- <br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
Guenter Waibel<br>
Berkeley Art Museum & Pacific Film Archive<br>
Digital Media Developer http://www.bampfa.berkeley.edu/>
Digital Imaging SIG Chair, MCN
http://www.mcn.edu/visig_subscribe.taf>
mailto:guenter@uclink4.berkeley.edu<br>
Phone <x-tab> </x-tab>510-643-8655<br>
Fax <x-tab>
</x-tab>510-642-4889<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~</div>
</body>
</html>
--============_-1220447449==_ma============--
</pre>
This archive was generated by hypermail 2b29
: Mon Jun 04 2001 - 12:21:50 CDT