Intro to Computer Vision
Lecture 13
Greg Shakhnarovich
May 13, 2010
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Guidelines and tips
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Community
Ask the community
Ask the community for help and clear up your study doubts
University Rankings
Discover the best universities in your country according to Docsity users
Free resources
Our save-the-student-ebooks!
Download our free guides on studying techniques, anxiety management strategies, and thesis advice from Docsity tutors
Computer Vision Lecture 13: Geometry and Stereo, Lecture notes of Computer Vision
A lecture note from a computer vision course, specifically lecture 13, covering the topics of geometry in two views, essential matrix and rectification, correlation-based stereo, window-based stereo, structured light stereo, range scanning, and stereo benchmarks. The lecture also discusses the evaluation of stereo methods and the importance of ground truth data.
Typology: Lecture notes
2011/2012
1 / 42
This page cannot be seen from the preview
Don't miss anything!
Related documents
Partial preview of the text
Download Computer Vision Lecture 13: Geometry and Stereo and more Lecture notes Computer Vision in PDF only on Docsity!
Intro to Computer Vision
Lecture 13
Greg Shakhnarovich
May 13, 2010
Review: geometry of two views
Figure: L. Lazebnik
Baseline - line containing OO′ Epipolar plane - any plane containing baseline Epipole - intersection of baseline with image plane Epipolar line - intersection of an epipolar plane with image plane
Correlation-based stereo
Source: A. Zisserman
Effect of window size: small large
Figures: Y. Boykov
Window-based stereo
Slide by L. Lazebnik
Evaluation of stereo
Need ground truth
How do we get it?
Don’t want to manually label each pixel...
Structured light “stereo”
From : Zhang et al. 2002
Idea: project known pattern of light on the scene Design light pattern so that it simplifies the correspondence problem Can use a single camera!
Structured light “stereo”
From : Zhang et al. 2002
Idea: project known pattern of light on the scene Design light pattern so that it simplifies the correspondence problem Can use a single camera!
Sharshtein & Szeliski
Range scanning
Alternative method for ground truth collection: range scanning
Technique: very precise version of structured light!
from Digital Michelangelo Project (Stanford)
Resolution: 0.29mm
Digital Michelangelo
Digital Michelangelo
model photo
Stereo: improvements?
What non-local constraints are we ignoring in a na´ıve window-based matching algorithm?
Multiple hypotheses could satisfy the epipolar constraint
Gee & Cipolla 1999
Uniqueness constraint
At most one match to each point in left image
Gee & Cipolla 1999
Ordering constraint
Point order on the surface of opaque object determines their order in both images
but...
Scanline stereo
Idea: try to match entire (epipolar) scanlines, enforcing contraints