-------- Original Message --------
Subject: Re: idea about image acquisition
Date: Tue, 20 Sep 2011 06:07:11 -0400
From: [email protected]
To: [email protected]
Dear listers!
I have already sent my recommendation to Darren Parsons(who made the
original request) to use a flatbed scanner, as we use for small fishes
(our paper in 2007-Cybium was mentioned in a previous email), but his
object size will definitely make problems for scanning on a desktop
flatbed scanner, although there are larger models available. Since the
discussion is going on, I would like to add a bit of my experience with
scanners. I will especially refer to two problems recently posted and
related to the nature of flatbed scanners. Most information may be
repetitive to our paper.
1.) Image distortion: I do not think that the position on the scanner is
important (like it is when objects are photographed through a curved
lens) since a flatbed scanner is expected to operate the same way over
its entire scanning surface, but z-axis-distortion on scanners seems
light-angle-dependent and only occurs along the x-axis (parallel to the
lamp and perpendicular to the scanning direction), and is scanner
specific (some scanners seem to have no distortion at all -
unfortunately I never found out which type it is). Distortion in
scanners is particularly a problem since it occurs in only one axis and
when the landmarks are not in the same plane on the object
(unfortunately the dream of a system of all landmarks in the same plane
does rarely come true in anatomically complex animals). So a potential
mathematical correction of the distortion is not so simple as you need
to know at what level along the z-axis a particular landmark is. The
simplest of course is to use this method only for small objects with a
rather compress shape and similar sizes in the dataset and ignore the
distortion since it will affect all the specimens the same way. The
worst you can do is to collect images from very different scanner
models. Then you will eventually find the scanner models explaining your
first or second principal component. We fooled around with such things
in a student course and created nice significant major PCs by for
example tilting the specimens a little bit (so exact positioning is
really important). The same happens if you scan the same set of somehwat
cylindrical objects parallel and perpendicular to the scanning direction
(needs to be strictly avoided). Then you end up with two artificial
populations.
@ David Thulman: I wonder about the error of 0.01 mm mentioned in a
previous email. To which ruler length and to what z-distance does this
refer? The espon scanners I tested some time ago yielded an error of
less than 1% at 4mm distance and up to 3% at 10mm distance (the maximum
depth of field/focus I recommend to be employed in image acquisition and
potentially already a problem in terms of distortion) from the scanner
plate.
2.) shadows: we tried a lot, but I actually had never big problems with
shadows (although I have to admit that fishes have quite a nice outline,
including fins). I usually use a white or light grey cover but a grey
background will also be attained when you move a white cover away from
the object (the background will darken with distance from the object and
the shadows may eventually disappear). I found differences between
scanners concerning shadows (a bit better in HP-scanners but those
yielded a worse image sharpness than epson) and stay with epson
perfection (4990 or V300) photo scanners for my small (less than 5 cm)
fishes now.
So, my call to the morphmet-family (or members who use scanners) is to
measure a series of graded rulers (e.g. in steps of 2 mm along the
z-axis, as we did for our paper, but since the distortion is linear, a
measurement at 0 and 10 mm z-distance would be sufficient) along the two
scanner axes and tell the community if you find a scanner model without
distortion on any axis (this model will then be probably sold out soon
...). Any better solution for the shadow- problem will be helpful as
well. Both solutions would for sure contribute to a widely used and very
cheap 2D-imaging method. Some of my colleagues and myself already travel
to field work (including expeditions) with a scanner in the luggage.
Cheers
Juergen
_
_
----------------------------------------------------------
<°))))><
Dr. Juergen Herler
Department of Integrative Zoology
Faculty of Life Sciences
University of Vienna
Althanstraße 14
A-1090 Vienna/Austria/Europe
Tel.: +43-4277-76313
e-mail: [email protected]
http://homepage.univie.ac.at/juergen.herler
