Event Details
Feature-based Image Matching and its Applications
Presenter: Dr. Radu Horaud - Senior Research Scientist, INRIA Rhone-Alps, Grenoble France
Supervisor:
Date: Fri, July 6, 2001
Time: 10:00:00 - 00:00:00
Place: EOW 430
ABSTRACT
Abstract
One of the most difficult problems associated with computer vision systems is the problem of image matching: Given several images of the same scene, find feature-to-feature correspondences between images. In this talk we present a method for characterizing image features such that the matching process tolerates large discrepancies between the various viewing conditions (large motions and/or large camera setting variations). Based on this characterization practical algorithms use multiview geometry and robust statistics in order to perform image registration. A number of examples illustrate this general purpose method, including image interpolation, video mosaicing, 3D reconstruction, and content-based image database query.