lecture4 - VisualSimulation CAP6938 Dr.HassanForoosh...

Info iconThis preview shows pages 1–12. Sign up to view the full content.

View Full Document Right Arrow Icon
    Visual Simulation CAP 6938 Dr. Hassan Foroosh  Dept. of Computer Science UCF © Copyright Hassan Foroosh 2002
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
    Today Last Lecture Multi-view modeling View Morphing Feature Tracking Good features to track (Shi and Tomasi paper) Tracking Structure from Motion Tomasi and Kanade Singular value decomposition Extensions
Background image of page 2
    Structure from Motion Reconstruct  Scene  geometry   Camera  motion Unknown Unknown camera camera viewpoints viewpoints
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
    Structure from Motion The SFM Problem Reconstruct scene  geometry  and camera  motion  from  two or more images Track 2D Features Estimate 3D Optimize (Bundle Adjust) Fit Surfaces SFM Pipeline
Background image of page 4
    Structure from Motion Step 1:  Track Features Detect good features corners, line segments Find correspondences between frames window-based correlation
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
    Structure from Motion Step 2:  Estimate Motion and Structure Simplified projection model, e.g.,   [Tomasi 92] 2 or 3 views at a time   [Hartley 00] [ ] n 2 1 f 2 1 f 2 1 X X X Π Π Π I I I = Images Motion Structure
Background image of page 6
    Structure from Motion Step 3:  Refine Estimates “Bundle adjustment” in photogrammetry Discussed in Bill Triggs reading Other iterative methods
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
    Structure from Motion Step 4:  Recover Surfaces Image-based triangulation   [Morris 00, Baillard 99] Silhouettes   [Fitzgibbon 98] Stereo   [Pollefeys 99] Poor mesh Good mesh Morris and Kanade, 2000 Morris and Kanade, 2000
Background image of page 8
    Feature Tracking Problem Find correspondence between  n  features in   images Issues What’s a feature? What does it mean to “correspond”? How can correspondence be reliably computed?
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
    Tracking Features Approach Compute motion of a small image patch Alternatively:  line segment, curve, shape
Background image of page 10
  Feature Correspondence Correspondence Problem Given feature patch F in frame  J , find best match in frame  I Solution Small displacement: Lukas-Kanade - = t y t x y y x y x x I I I I v u I I I I I I 2 2 Large displacement: discrete search over (u,v) – Choose match that minimizes SSD (or normalized correlation) When will this fail? Find displacement (u,v) that minimizes SSD error over feature region
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 12
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 32

lecture4 - VisualSimulation CAP6938 Dr.HassanForoosh...

This preview shows document pages 1 - 12. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online