VASC Seminar Announcement
            =========================

Date: Monday, 12/6/99
Time: 3:30-4:30pm
Place: Smith Hall 2nd Floor Common Area

Speaker:  Jianbo Shi
          CMU Robotics Institute
          http://www.cs.cmu.edu/~jshi

Title:  A computational model of grouping and image segmentation

Abstract:

Grouping is a complex problem with many intertwined issues and subgoals.
In this talk, I will address some of the issues of grouping in the
context of image segmentation.

We view grouping as a process of 1) encoding visual similarity between
the image primitives, and 2) extracting groups in the image based on
these pairwise similarity measure. The proposed grouping engine, 
Normalized Cuts, formulates grouping as a hierarchical graph partition 
problem.   From the input image (or image sequence), a weighted graph 
is first constructed base on local image feature similarity.  The graph 
is then recursive sub-divided according to a global partitioning criteria 
called normalized cut.  The normalized cut criteria measures both the 
total dissimilarity between the different groups as well as the total 
similarity within the groups.  An efficient algorithm is developed for 
finding optimal partition by solving a generalized eigenvector system.

We have applied this algorithm to image segmentation based on
color/brightness, texture and motion cues.

This is joint work with Jitendra Malik.