Martin Szummer

I have moved from the MIT Media lab to the MIT AI lab. This page describes my previous Media lab research, my new page is here.

Ph.D. candidate in Electrical Engineering and Computer Science, Area II

Research interests

Image processing: image query, color, movie processing
Machine Learning, Machine Vision

Advisor: Prof. Aaron Bobick

Education

BS, Mathematics
Massachusetts Institute of Technology, 1994

BS, Computer Science
Massachusetts Institute of Technology, 1995

M.Eng. Electrical Engineering and Computer Science
Massachusetts Institute of Technology, 1995

Experience

Natural language processing; Computational Geometry (IBM Yorktown)
People detection (Daimler-Benz, Ulm, Germany)
Medical image databases (University of Erlangen, Germany)
MIMD Parallel Computing (MIT AI lab)
Automated theorem proving (MIT AI lab)

Research Description

In the future, huge databases of movies will be available. I want to make tools that can help in the retrieval of specific image sequences from such a database. As a first step, I have written a program that detects scene changes by comparing color histograms. Now, I would like to extend single image query techniques to image sequences. I would like to detect specific moving textures, such as leaves fluttering in the wind, or waves on the sea. Even more general, I would also like to find image sequences similar to a given image sequence. This work should take into account human perceptual attention--what is it that we notice and remember?

Publications

Indoor-Outdoor Image Classification

Martin Szummer and Rosalind W. Picard
IEEE International Workshop on Content-based Access of Image and Video Databases, in conjunction with ICCV'98. Bombay, India
Also appeared as MIT Media Lab Perceptual Computing TR #445.
Abstract and compressed Postscript file (10 pages, 1100 Kb, no color),
alternatively color Postscript file (requires Postscript level 2) (not available yet).
Color images in paper

Temporal Texture Modeling

Martin Szummer and Rosalind W. Picard
IEEE International Conference on Image Processing (ICIP 1996). Lausanne, Switzerland.
Also appeared as MIT Media Lab Perceptual Computing TR #381.
Abstract and compressed Postscript file (4 pages, 412 Kb),
alternatively Adobe Acrobat PDF file (500 Kb)
Image Sequences of Temporal Textures

Temporal Texture Modeling

Martin Szummer
M.Eng. Thesis, MIT, Sept 1995.
Also appeared as MIT Media Lab Perceptual Computing TR #346
Abstract and compressed Postscript file (56 pages, 1.3 Mb)
Image Sequences of Temporal Textures

Modeling user subjectivity in image libraries

Rosalind W. Picard, Thomas P. Minka, and Martin Szummer
IEEE International Conference on Image Processing (ICIP 1996). Lausanne, Switzerland.
Also appeared as MIT Media Lab Perceptual Computing TR #382.
Abstract and compressed color Postscript file (4 pages, 767 Kb),
alternatively Adobe Acrobat PDF file (304 Kb).

Learning image browser

A hierarchical image database browser that learns from user interaction.

NL Understanding with a Grammar of Constructions

Wlodek Zadrozny, Marcin Szummer, Stanislaw Jarecki, David E. Johnson, Leora Morgenstern
Proc. Coling'94, Kyoto, Japan, 1994
Abstract and compressed Postscript file (5 pages, 36 Kb)
Conference site.

Interests

Swimming, Skiing, Ballroom dancing, Tennis, Squash, Piano

Birthplace

Warsaw, Poland

Favorite book

J R R Tolkien. The Lord of the Rings

Most wishes to have dinner with

chopsticks!

Personal Web Page

Vismod Home Page