I have two 1bit-images of snowflakes.
These two images are taken shortly one after the other of the same group of
snowflakes, but they moved a little in the meantime in different directions
or spinned, so they don't look exactly the same. I want to "match" the
snowflakes of these two images.
This has to be done with hundredthousends of such couples of images.
What I am doing so far is:
I am parametrizing each snowflake with a few variables (such as lenght,
width, perimeter, coordinates of focal point etc.). I am writing each
variable in a vector, so each snowflake is represented by a vector. This is
done by WIT (Visual Programming Software from Logical Vision).
What I have to do now is to match these vectors to find out, which snowflake
in one image is belonging to which snowflake in the other image. The idea is
to measure distances (in a C++-program) between all couples of vectors and
to minimize the sum of all distances. Vectors with a minimized distance
would represent matching snowflakes.
The problem is similar to a fit-algorithm. How do I walk fast through all
couples of vectors and find the corresponding ones?
Examples in any programming-language are welcome!
Did someone of you have a similar problem, for examle with matching moving
viruses under the microscope?
Do you know a link? Which is the best newsgroup for my question?
Do you have a general idea?
Any help would be very appreciated!
Raphael Schefold
