We should start, welcome to the Tuesday session.
Unfortunately we have to have a shorter session today
because I have to leave at 12.30 for a talk,
a very important talk.
And that means that I don't want to summarize today
where we are, but that should not encourage you
to answer the question in the evaluation
is there a storyline or not in the wrong manner.
We have a storyline, okay?
So yesterday we started to look into the problem
of image registration and in this context we said,
okay, let's look at the following problem.
It's very easy to characterize.
We have points and we have points in the second image
and these points are rotated and translated in plane.
So we have a 2D rigid transformation,
no deformation at all.
And the question was how can we compute the rotation
and translation out of these point correspondences?
This is a trivial problem, by the way.
This is not something where anybody is very excited
if you present a solution.
What we found out yesterday is that our knowledge
about complex numbers helps a little bit to describe
and to formalize the problem in a way that we end up
with a system of linear equations.
There are also other ways to do that.
This is not the only way.
But I personally find this pretty nice
and we do the following.
We say these points, the 2D coordinates
are considered as complex numbers.
The X value is the real value,
the Y value is the imaginary value.
And then we look what happens, I mean,
if I rotate and translate a point.
Basically, I can say complex number
representing the point PK1, PK2
is generated out of the point QK1, QK2
by a multiplication with a rotation complex number
and by the intercept of a translation complex number.
And the reason why we can do this here is
we looked at the algebraic operation of multiplication in C.
Not C, yeah?
And it turns out that we have a very nice geometric
interpretation of multiplications of complex numbers.
Multiplication of two complex numbers
is basically nothing else but a rotation and scaling.
And if the complex number has the norm one,
we know this is just a rotation and no scaling.
Presenters
Zugänglich über
Offener Zugang
Dauer
01:01:51 Min
Aufnahmedatum
2011-01-25
Hochgeladen am
2011-04-11 13:53:29
Sprache
de-DE