Michael Werman - computer scientist

"I see related: When I see three people talking, I think of a graph with edges in between them."
(links for Michael Werman)

What kind of role pictures play in your professional field?

My whole life I spend trying to understand pictures.

My whole life I spent trying to understand pictures, improve pictures, enhance pictures, make pictures. I am a computer vision researcher: For me pictures are numbers, a square of numbers or a cube of numbers - maybe it’s a movie.

How has your conception of images changed ? Because you grew up also with analogue images.

I still see analogue images

I still see analogue images. If I see an image many times I think of how to turn it into another image or how to get it into the computer as a description.

Did this change your experience of images?

Not always, but I think sometimes. Yesterday we talked about that word drawing.

Refering to Walter Bejamin the art curator yesterday talked about the gesture of scratching, which means to take some material away.

drawing in computer science

Drawing in English is different. It means you draw water out of the well. The same way I am looking at the world: how can I take it out of the world and put it in the computer.

Did the web change your experience of images?

you can't get virtual reality feeling on the web

Not in images for me. A long time before the web existed, since I was a student I had pictures on the computer. I once started working on an online store project. The problem is that you still can’t get a virtual reality feeling on the web. In virtual reality we can walk through a store, a park or a stadium. But it does not work yet on the web, because the perspective is too close.

You talked about converting images into a description. So you formalize them by the means of mathematics. Do images exist in pure mathematics?

The proving things can’t be done with pictures as they are done in the end with symbols.

There are two main types of mathematicians: the geometry kind of people who think in pictures and the algebra kind of people who are logicians and think in symbols. The big boom of the last hundred years with mathematics to my sorrow was the algebra one: They turned geometry into algebra.
So there are different kinds of people. Most mathematicians even though they work in thousand dimensions draw two dimensional pictures to get the idea. But there are still mathematicians who work completely symbolically.
The proving things can’t be done with pictures as they are done in the end with symbols. You can fool yourself with pictures, but not so much with symbols, because it’s a more mechanical kind of way looking at things.

We talked this afternoon about images and you mentioned that images have a lot of information. How does it always fit if you describe images by such formalized codes? Is there all the information of the image in the code?

My main field of research is to take big pictures with lots of information and get a one line description.

very easy answer: "cancer yes" or "cancer no"

What I always try to tell my students is that you are often looking only for one bit of information. In an hour of a video there are tons and megabytes of information, but you are looking for one little thing. You see a movie for hours and all that has left is there was "a beautiful girl on the screen". That’s five words.
In a medical image you want to look in a bunch of slices of a human being, but the answer is very easy: “yes, cancer” or “no cancer.” The information is very big, however the question is which answer of the two is the right one.
My main field of research is to take big pictures with lots and lots of information and get a one line description.

We talked about the language of mathematics. For non-mathematicians the factor infinity is always a little bit mysterious. How do you treat the infinite?

But what you really do with infinite is problematic.

Infinity was never a problem for me. But what you really do with infinite is problematic. You can show that you take a ball and cut it into five pieces. If you put these five pieces together, the ball is twice as big as the original ball with no holes in it. You can’t imagine what I mean.

No.

Thinking about infinity seems all right but a lot of things can be wrong.

Of course you can’t. If we use symbolic rules taking that axioms of infinity you can do that. Thinking about infinity seems all right but a lot of things can be wrong.

A further question is if there are some results where you feel the math?

I see related.

Yes there are. People often feel that I am thinking in mathematical metaphors. I see related: When I see three people talking, I think of a graph with edges in between them.

Do you feel the maths in architecture, e.g. in ornaments, in mosaics?

I like more fluent, more continuous, less structured mathematics.

For me that’s group theory. I like more fluent, more continuous, less structured mathematics. In mountains, moving water and moving clouds I see the equations.

Do you see a relation between abstract images and music?

I don’t like the mathematical in music, I feel a kind of constraint.

I don’t put them together. If I hear music and I close my eyes, then I see colors and pictures, but not mathematics. Some musicians are very mathematical, like Bach. But I don’t like that in music, I feel a kind of constraint.
But actually I read a book that is called “Hackers and Painters” which was trying to mix those things together.

How did you come to work with pictures?

It was just fluke. I did Masters in cryptography and then got married and moved to a different city. It just worked out that way.

But is has something fascinating for you?

movement amazes me: Even Even if you see a person walking from a hundred meters away, you can say who it is.

Yes. What I do is very simple. I think my visual memory is not very good. But one thing I can see and which always amazes me is: You normally understand movement, but the way people move is really, very personal. Even if you see a person walking from a hundred meters away, you can say who it is. Movement is a great thing.

How do you treat this time factor?

We are just starting to understand movement.

The tools to follow and figure out things in images and movies as trajectories in space/time are still fairly poor. For instance you can track things, so you look for similar things in different images. I have been trying to do this for a long time, but it is still not perfect. We are just starting to understand movement.

A very personal question: What is your favorite image at home?

simple flowing shapes

A Kandinsky - an abstract picture with simple flowing shapes, no straight lines and smooth colors.

interview from 1st February 2005 at Schloß Dagstuhl

Links for Michael Werman:
Michael Werman/ Institute of Computer Science, The Hebrew University of Jerusalem

Publications:
(with Isaac David Guedalia, Mickey London)