dcasadon at magnus.acs.ohio-state.edu (Donald J Casadonte) writes:
: As far as I can tell from your description, Dr. Travis, it seems we are
: doings some of
: the same things with the system we have here (a Tracor 8502 Image Analysis
: System (Tracor is now known as Noran, I believe)). I am looking at the
: distribution of cell cross-sectional lengths instead of cell wall thickness
: (although, that would not be such a bad idea...). For anyone who wants to see
: a rather pretty, but meaningless, use of image analysis Science News (Dec., 14,
: 1992) published a palleted (i.e., colorized) version of the optical density
: transmission through a clarinet reed which I did last year. This type of
: colorization can be used to show the scraping patterns of musicians who "work
: on" their reeds (such as oboe and bassoon players).
I'm not sure what you mean by "cross-sectional lengths" ?
Are you using a sampling method to estimate the distance between the
lumens of adjacent cells or are you measuring the distance between
points on the image selected by a human observer?
We compared our automatic method with manual estimates on the same
images and found them to be in reasonably good agreement. The point
about doing it automatically is that we can investigate in more detail
because the work is less manually intensive.
: I am currently doing something very exciting from a visualization
: standpoint-I am sectioning a clarinet reed on a confocal microscope (which uses
: a laser light source) and then dumping the information to a Silicon Graphics
: program which will reconstruct the reed interior three-dimensionally.
Would that by any chance be a Molecular Dynamics Confocal system? We've
seen the spec. of an SG-based system from them and it looks quite
impressive - what is your impression of the system, Donald?
: That these types of techiniques are gaining momentum in biology might be
: reflected in the fact that the medical school at Ohio State has just concluded
: a special seminar showing off its imaging equipment.
: Hope this is of some use in showing a few of the many possibilities of
: image analysis in plant science.
Quite a lot of work has already been done on the use of image analysis
in plant science. For example, I read some very impressive work by
Tarbell, Tcheng and Reid (1991) Trans. American Soc. Agricultural
Engineers, 34, 2264-2271. They have used inductive learning techniques
in conjunction with image analysis to examine the growth of maize
Unfortunately, a lot of image analysis systems are 'turnkey' systems
and more or less dedicated to the task that the manufacturer intended
unless you can find the (large amount of) money to have a new 'custom'
application written for the system. This is not a problem if the task
is narrowly defined and there is little need for change, but in the
research environment that many of us work in a 'turnkey' system is not
It is very common, however, to seriously underestimate how difficult
some image analysis tasks are because they are trivially easy to do
with our own human visual system. A reasonable compromise that many
people use is to write their own programs using a commercially obtained
image analysis library which lets you 'buy-in' image analysis skills
without sacrificing flexibility. There are also many image analysis
packages available on the net. The comp.ai.vision news group and
archive at ftp.ads.com are good places to start looking if anyone is
Tony Travis <ajt at uk.ac.sari.rri> | Dr. A.J.Travis
| Rowett Research Institute,
| Greenburn Road, Bucksburn, Aberdeen,
| AB2 9SB. UK. tel 0224-712751