"Nick" <lifanshi at usc.edu> wrote in message
news:3e5daba2.21019734 at news.usc.edu...
| I have to deal with the neuron images from the confocal microscope.
| was supposed to measure the movement of the neurons; however, the
| slide moved too. Therefore, the image registration techniques is
| needed to eliminate the movement of the slides from the photos.
|| There are many softwares for image registration or alignment, but
| of them work! I would like to know if there are any software that
| solve this particular situation, otherwise if there is an algorithm
| solve it. Thanx!
|| The idea to solve it seems to be simple. Calculating the maximum
| movement and delete it. However, it is really hard to do it in the
I've never done it myself [except in mind's 'eye'], but I've thought
about it a lot :-]
If you work with 'time'-sequenced photographic images of the neuron,
or a portion of it, this problem goes away, at least with respect to
relatively small movements of the slide. Just keep the same settings
on the camera over all photos. Then analyze the images.
You can scan your images, then overlay them in a CAD program's
layers. 'Register' the images which is relatively-easy because they
are 2-D static. You can even write a simple macro in such CAD
software that'll step through such overlaid layers which will give
you a a 'time'-sequenced animation of the cell's intrinsic motion.
Also, CAD programs have their own surface-area analyses capabilities
built right in.
I understand that there's a lot of software out there expressly
designed to address this particular problem. Seen some of it in
action when I've visited labs and universities, but I'd approach it
as above, with 'time'-sequenced images and software that allows me to
program what it is that I want to do [which doesn't try to 'think'
for me - it doesn't know what I want to do :-]
The main thing is to have a long-term approach that obviously works,
and the dedication to stick with it, and not skip steps. You know -
grind it out.