Okay, so welcome to the Tuesday afternoon session, 45 minutes, exciting topics about

interventional medical image processing.

And currently we are in the stage of preprocessing.

So how do we need to preprocess the images such that they have a higher value during

the procedure than before or than originally acquired?

And one step towards this goal is to compute distinguished points in the image like edges

or corners or things like that.

And the other thing is that we want to compute straight lines.

That's the task we are currently considering.

Find structures that are separated by straight lines and how can we build a straight line

detector using the image information that is available.

And I have shown to you one motivation which makes hopefully clear why this is important.

The detection of straight lines is necessary to find here a proper collimation during procedures

to reduce the dose that is applied.

I mean if you get a single exposition that's usually not a big deal.

I mean the dose is rather limited.

If you get a surgery and the doctor requires during the surgery hundreds of x-ray images

then it's important that the dose is reduced to minimum such that only those parts of the

human body are exhibited to the x-ray system that is required to be visible in the projection.

And for these purposes they use here this light interface where they see basically the

area that is exhibited to the x-rays and it can be adjusted properly.

And where do I have here?

And this is a typical example.

You see here the skull, the head.

Here you see the spine in this area, in the neck area.

And here you see the shoulder and you're just interested in the spine.

So if you do a surgery of the spine and you want to reduce the amount of blood you generate

of course you use a lot of image information to make the cuts as small as possible.

And what doctors need for instance is a rectification of these images that we cut out this image

that is just showing the interesting part and then we rectify it and leave the background

where it is.

Just throw it away.

And the question is how can we detect straight lines?

And that's a problem that we have started to consider yesterday.

Where is my English teacher?

Here.

Where we have started to consider this yesterday.

And we have seen different ways to compute gradients.

We also know about the role of gradients.

If I compute for a certain image point, a gradient, I get a certain amount of information

and this type of information tells me what is the strength of the edge where I'm currently

sitting on and what is the direction of the maximum change in the mountains of intensity

values.

Okay?

So images for us are basically a plate with mountains on it.

And we will grow in different ways of illustrating images in terms of geometry and functions.

So that's what we currently need.

That's a plate.

And we have here the mountains on the plate and we compute the gradient on certain positions.

How can we compute the gradient?