Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
C4B Computer VisionMichaelmas Term 2009
Andrew Zisserman
Lecture notes and more at:
http://www.robots.ox.ac.uk/~az/lectures/cv
Course summary
1. Introduction; imaging geometry; camera geometry
2. Salient feature detection – edges, lines, corners
3. Recovering 3D from two images 1: epipolar geometry
4. Recovering 3D from two images 2: stereo correspondence
5. Structure and motion 1: estimating the F matrix, RANSAC
6. Structure and motion 2: SIFT, estimating H, more than two views
7. Visual motion and tracking
8. Object detection and recognition
Structure and Motion
Reconstruct • Scene geometry
• Camera motion
Unknowncameraviewpoints
Structure and Motion from Discrete Views
• Introduction
• Computing the fundamental matrix, F, from corner correspondences
• Feature matching
• RANSAC
• Estimation
• Determining ego-motion from F
• SIFT for wide baseline matching
• Computing a homography, H, from corner correspondences
• More than two views
• Batch and sequential solutions
Example
Original sequence Camera track and 3Dpoints
Example: compute F from 8 point correspondences
Images from a parallel camera stereo rig – epipolar lines y = y/
2
2-2
-2
-4 4
x1x2
x3
2
2-2
-2
-4 4
x/2
x/3
x/1
just consider first three points
write f in matrix form as
1