Message-Id: <mailto:199412161609.KAA13948@library.wustl.edu> Date: Fri, 16 Dec 1994 11:06:21 EST From: mailto:rarosenb@GANDALF.RUTGERS.EDU> Subject: Re: Kontron camera To: Multiple recipients of list IMAGELIB
In response to Victoria Yturralde's query about the Kontron camera, here is
an abridged report of the results of our test of the Kontron in January
1994. We (the Edison Papers) were one of five sites testing the camera.
There is a discussion list about the camera, originally centered on these
tests, at mailto:KONTRON@U.WASHINGTON.EDU (you can get the archive from the
listserver itself, or ask the list owner, Mary
mailto:Keeler--MKEELER@U.WASHINGTON.EDU). And I would recommend contacting Roche,
the American distributor of the camera, to find one near you that you could
observe. I understand that image capture is now considerably faster than it
was a year ago, which was even then a little faster than an equivalent
scanner (16 seconds for a 24-bit image at 3k x 2k pixels).
If there is unintelligible technospeak in here (like that last
parenthetical remark), I will be glad to clarify it (if I can). I also
excised the description of the Edison Papers and our plans, which seemed
peripheral to the question at hand.
***********
Our system, which was supplied entire, worked flawlessly, and it couldn't
have been much simpler. Check position, lighting intensity, focus,
resolution, white balance if desired; preview; capture. Our supplier spent
a couple of hours the Thursday before the test showing me how to use the
software (equipment setup was trivial). The camera had a 28mm (26mm?) lens.
We used WinCam for capture and processing and Adobe Photoshop for
post-processing fooling around. We saved images as TIFF files and
experimented with JPEG. We ran on a 50MHz 486 with 32Mb RAM, a 24-bit SVGA
video board, and a 14" monitor. We used two 250-watt 3200K bulbs for
lighting, which gave plenty of illumination and fine color, although we were
not worrying about professional-level lighting and ignored shadows and hot
spots on the artifacts. (In our printed volumes we use photographs of
artifacts, and I have watched a professional take 45 minutes to set up a
shot of an extremely placid telegraph instrument. There's no reason to
expect lighting requirements to change with a digital camera, so we let it
go.) Our only problems were space, as we had "just" 200Mb for images, which
meant that we couldn't keep more than a couple of full-resolution color
images [21Mb each], and RAM, which sometimes jammed up when WinCam and
Photoshop were both open.
In a nutshell, I would say that our results were spectacular. We did some
testing in 8-bit grayscale and had no complaints. In fact, on much of our
material, which is largely pen, pencil, or type on (erstwhile) white paper,
it delivered a sharper image than 24-bit color just because it used only the
green light and sensors in the camera. It is faster, too, taking about half
the time of an equal-resolution color scan, and yields an image file
one-third the size.
We mostly tested color images, though. On artifacts, as I think everyone
found, the images were remarkable even at half-resolution and at full res
they showed details barely discernible to the unaided eye. Shooting
documents in color allowed striking enhancement in Photoshop, so that green
ink that had faded almost to invisibility could be read far more easily than
on the original, and variously colored bleedthrough on onionskin could be
eliminated. One benefit of such malleability is that time-consuming
adjustments during shooting, although much easier than with a microfilm
camera, are less necessary, since the user will be able to "fix" difficult
images in ways impossible on film.
If you want a test of the camera's clarity that will stun visitors, take a
ten or twenty dollar bill of recent vintage (it has to be 1990 or later) and
shoot it, with the frame width equal to the bill's, full resolution. Then
blow it up and look at the outside line around the central portrait (Jackson
or Hamilton). You will see that the "line" is in fact not a line at all.
It (the third line around Jackson or the second around Hamilton) is in fact
"THE UNITED STATES OF AMERICA" printed repeatedly in very tiny letters.
We experimented with different resolutions, finding half and even one-third
resolution usually fine for the documents. A full-res grayscale scan is
about 7Mb. We used JPEG to squeeze files down and were hard-put to see
changes in images (full-color or gray) until we compressed beyond 15 or
20:1. We intend to compress about 10:1 if our long-range project gets off
the ground. At that ratio we could see no degradation.