Perspective Projection Properly Models Image Formation
Speaker: Berthold K.P. Horn , CSAIL
Date: March 10 2004
Time: 1:00PM to 2:00PM
Location: NE43-8th Playroom
Contact: Jake Beal, x3-6524, jakebeal@mit.edu
Relevant URL:
Abstract:
Methods based on projective geometry have become popular in machine vision because they lead to elegant mathematics, and easy-to-solve linear equations. It is often not realized that one pays a heavy price for this convenience.
Such methods do not correctly model the physics of image formation, require more correspondences, and are considerably more sensitive to measurement error than methods based on true perspective projection. In this talk we find that for the example of exterior orientation:
(i) Methods based on projective geometry are fundamentally different from methods based on perspective projection;
(ii) Methods based on projective geometry yield a transformation matrix T that in general does not correspond to a physical imaging situation that is, a rotation, translation and perspective projection;
(iii) Optimization methods based on the real physical imaging equations (true perspective projection) produce considerably more accurate results.
See other events that are part of Dangerous Ideas Seminar Series Spring 2004
See other events happening in March 2004
