MIT Department of Electrical Engineering & Computer Science
Curvature Flows for Image Segmentation, Smoothing, and Enhancement
Anthony Yezzi
University of Minnesota
Thursday, May 1, 1997
4:15 PM (4:00 refreshments)
RLE Conference Room, Room 36-428
EECS Special Seminar
Abstract
In this talk, we discuss the application of partial differential
equations defined by curvature-driven flows to several key
problems in two and three dimensional image processing and
computer vision. In particular, we will present our new active
contour method which unifies the curve evolution approach for
geometric contours and the classical energy method for snakes.
In our technique, the final contour will yield a geodesic (or
minimal surface in the 3D case) with respect to a conformally
Euclidean metric chosen to highlight features of interest (edges,
textures, etc.) in the image.
We will also consider related curvature flow models, projected
along directions which tend to preserve edges, in order to
develop a general image-enhancement filter. The essential idea is
to treat an n-dimensional m-vector valued image as a surface
embedded in (n+m)-dimensional Euclidean space. From this general
filter, we derive special filters for the cases of 2D and 3D
greyscale as well as color imagery.
We will highlight the application of our methodology to medical
imagery from a variety of modalities: ultrasound, MR, CT, and
nuclear.
URL of this page:
http://www-eecs.mit.edu/AY96-97/events/28.html
Created: Mar 20, 1997
|
Modified: Jun 24, 1997
This announcement is from the MIT EECS 1996-97 archive.
|
Current events
To MIT EECS home page
|
Your comments
and inquiries are welcome.
