So, welcome to the morning session.

We are currently in the field of defect pixel interpolation.

And before we continue discussing our frequency-based interpolation scheme, the second one, actually,

I would like to summarize what we have considered so far and what are the major messages we

have transferred to you so far.

So this semester, we talk about diagnostic medical image processing.

And this is a cloud of diagnostic medical image processing.

And what we did is, first of all, we looked at different modalities, but that was a very

rough and brief overview.

If you are interested in how these modalities work, how these modalities are used to do

a proper diagnosis, how these systems make use of physics, then you should attend lectures

on medical imaging.

We offer several lectures in this field.

For instance, in the EE department, there is a lecture on medical imaging that is given

by a Siemens employee.

Then we have usually some block courses on medical imaging and algorithms.

So we have several offers in this field.

And please accept that I just gave a very hand-waving overview over these things.

This is not the core topic of this class.

Our topic is basically, once we have activated our sensor and we have acquired a signal,

what can I do actually with the digital data?

That's the problem we are considering.

And we started out with one chapter on acquisition-specific preprocessing.

And I have to say that we distinguish between preprocessing and post-processing.

This is some kind of artificial classification.

This is also not standard in the literature.

But on the other hand, it makes sense because we look at two different stages of image processing

or signal processing.

With respect to preprocessing, we just mean the link between the sensor and the monitor.

So we capture the signal using a sensor, like a flat panel detector in X-ray imaging.

We acquire a signal.

We transform the signal into a digital signal.

And then the digital signal is sent to the monitor, to the graphics card and then to

the monitor.

And on the road to the monitor, we can do some image processing.

And everything we do in between the sensor and the monitor is preprocessing.

Once the data are stored in the database, the doctor brings up again the data and does

some post-processing, like measuring or segmentation, finding the contour line of the liver, finding

the bones, doing bone segmentation and things like that.

This is called post-processing in this lecture.

I have no idea whether this is used at any other place in the world.

But actually, that doesn't matter.

For us, it's important to know that there is a difference because from the detector

to the monitor, usually we have certain constraints saying we have a limited time frame.

And another very nice constraint is that basically the FDA is not looking into the details.

And whatever you do in between the detector and the sensor, this is just the signal that

is measured, that is shown on the monitor.

If you do some smoothing, the doctor has the feeling this is the signal I have measured.

If you do post-processing, they will say you did some computations.

This is no longer the original signal.