Good morning everybody.

So welcome to the Tuesday session as usual.

Let's consider the big picture and then we will continue

in the text considering probabilistic models

for inhomogeneity correction in MR imaging.

So the lecture in winter semester is basically

about image processing methods

for diagnostic applications in medicine.

And we know now a lot about modalities.

What I'm doing here.

About modalities for medical imaging.

And then we started out to do pre-processing.

And the pre-processing chapter was basically

on methods that can be applied on the way

from the detector, from the sensor to the monitor.

That was the way we defined pre-processing.

And we said, okay, let's look at the principles

that we borrow from physics to acquire images.

And then let's think about what happens

in terms of artifacts in this context.

And how can we in this context

and how can we correct for these artifacts.

And we have seen very nice applications for x-ray imaging.

There are just two to mention.

We considered image intensifier technology

and flat panel detectors.

Very exciting applications.

But at the end of the day,

we discussed the basics of parameter estimation.

We discussed basics of parametric modeling of mappings.

We discussed basics of least square estimates.

We also learned about the singular value decomposition.

We also have seen that we can use the Fourier transform

and the convolution theorem

to get rid of the defect pixels.

So we have learned a lot of interesting stuff.

And in the second part, we considered MR inhomogeneities.

That is a low frequency distortion of intensities.

And we have seen a bunch of methods

to do the correction of it.

That's basically where we are in.

And today we will start with 3D reconstruction.

And that means we will look into the problem.

What can I do if I have multiple images

of one and the same scene?

Can I generate higher dimensional information?

Can I compute, computational engineers, listen.

Can I compute out of the projection images

higher dimensional information?

And the answer is yes.