Research work with Micha Lindenbaum

My research is in computer vision* and other image related issues, but especially in object recognition* and grouping*. I am interested both in theory for predicting the feasibility and the performance of such tasks, and in building actual systems for carrying them out.

Master's and Ph.D. theses require both theory and practice. Students who want to work on topics in computer vision should not be afraid of simple mathmatics and should have reasonable programming skills (e.g., MATAM or Data Structures courses). Before starting to work on a thesis, a student should take at least one course in image processing or vision. The cooperation usually starts with reading and discussing some papers, or with a thesis related project. Such projects may be done also at the undergraduate level.

Students are expected to be clever, to suggest a lot of ideas, and to take responsibility for shaping their theses, rather than just doing what they are told. They are also expected to develop a broad interest in computer vision and to play an active role in the Intelligent Systems Lab, leading projects, contributing ideas for courses, and so forth.

Some useful links for students are:

The USC bibliographic list for computer vision--you can find almost everything you need here. Google scholar is another useful source.
Scientific documents in science and engineering are usually written in LaTex.
Many lectures on image related issues are given in the Pixel Club , by students, lecturers, and guests. The lectures are usually on Tuesdays 11:30 -12:30, so reserve this slot. The Pixel is managed only by students. To be on the mailing list where pixel lectures and other computer vision lectures are announced, send mail to the moderator.
The experimental activities take place in the Intelligent Systems Lab, where equipment and advice are available for work in vision, robotics, graphics and AI.

* Computer vision-A collection of techniques for "understanding" the world from its pictures, with the aid of a computer program.

* Object recognition-Finding particular objects in scenes from images of these scenes.

* Grouping-Dividing images into (not necessarily disjoint) groups in some semantically meaningful way. A partition where every object corresponds to an object is the common goal.