DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

UNIVERSITY OF CALIFORNIA, SAN DIEGO

Instructor: Serge Belongie, Assistant Professor, EBU3B 4118. Office Hours: TuW 3:30-4:30pm.

Teaching Assitant: Satya P. Mallick, EBU3B Basement 240A. Office Hours: M Fr 2:00-3:00pm

Note: when emailing the instructor with questions about the class, please put "cse252b" in the subject line or use the following address:

sjb+cse252b | cs.ucsd.edu |

Class section id for CSE252B: #559463. Lecture: TuTh 2:00-3:20pm, Center Hall 205. Class mailing list: http://graphics.ucsd.edu/mailman/listinfo/cse252b.

Topics to be Covered: Geometric models of image formation. Projective Geometry. Interest point detection. Robust feature matching across wide baselines. Calibrated and uncalibrated reconstruction of 3D models of objects from 2D images. Construction of photo-mosaics. Camera calibration. Optimal camera pose and structure estimation.

Prerequisites: linear algebra, calculus, probability and statistics.
This course makes extensive use of Matlab. Click here for
information on Matlab. Assignments should be prepared using
**LaTeX**. If you are not familiar with **LaTeX**, click here.

Handouts/Readings:

- Introducing Matlab [html]
- MaSKS Ch. 1-3
- HZ Sec. 2.0-2.4 [pdf]
**Assignment #1**[pdf]- MaSKS Sec. 5.1, 5.2.1-5.2.2, 5.3
- A computer algorithm for reconstructing a scene from two projections (
*Longuet-Higgins*) [pdf] - In Defense of the Eight Point Algorithm (
*Hartley*) [pdf] - Perspective Transform Estimation (
*Wren*) [html] - A Plane Measuring Device (
*Criminisi, Reid and Zisserman*) [html] [ps.zip] - Algorithms for Plane-Based Pose Estimation (
*Sturm*) [pdf] **Assignment #2**[pdf] [files]- Metric Rectification for Perspective Images of Planes (
*Liebowitz and Zisserman*) [pdf] - Autocalibration from Planar Scenes (
*Triggs*) [pdf] - MaSKS Ch. 6 and Appendix 6.A
- HZ Sec. 2.7, 3.2.1, 3.2.3, 3.5-3.7, 8.5 [pdf]
- Motion and structure from point correspondences with error estimation: planar surfaces (
*Weng, Ahuja and Huang*) [pdf] **Assignment #3**[pdf] [files]- Self Calibration from Multiple Views with a Rotating Camera (
*Hartley*) [pdf] - Self-calibration of rotating and zooming cameras (
*de Agapito, Hayman and Reid*) [pdf] - MaSKS Sec. 5.2.3
- MaSKS Ch. 4
- A Fast Operator for Detection and Precise Location of Distinct Points, Corners and Centres of Circular Features (
*Förstner and Gülch*) [pdf] - Junction detection with automatic selection of detection scales and localization scales (
*Lindeberg*) [ps.Z] - Feature Based Methods for Structure and Motion Estimation (
*Torr and Zisserman*) [pdf] - The Development and Comparison of Robust Methods for Estimating the Fundamental Matrix (
*Torr and Murray*) [pdf] **Assignment #4**[pdf] [files] [bonus photos]- SIFT Keypoint Detector (
*Lowe*) [website] [pdf] - MSER paper (
*Matas et al.*) [pdf] - An iterative image registration technique with an application to stereo vision (
*Lucas and Kanade*) [pdf] - Good Features to Track (
*Shi and Tomasi*) [ps] [pdf] Tech Report [ps] - Making Good Features to Track Better (
*Tommasini et al.*) [pdf] - Hitchcock Zoom effect (
*Wikipedia*) [www] - Shape and Motion from Image Streams: A Factorization Method (
*Tomasi and Kanade*) [pdf] - Affine Structure from Motion (
*Koenderink and van Doorn*) [pdf] **Final Exam**[pdf]

Lecture Topics:

- Lecture 1 (Apr. 4): Geometry of Image Formation, Homogeneous Coordinates
- Lecture 2 (Apr. 6): Homogeneous Linear Least Squares Problems, Two View Geometry
- Lecture 3 (Apr. 11): Epipolar Geometry, Calibrated Reconstruction
- Lecture 4 (Apr. 13): Planar Scenes and Homography
- Lecture 5 (Apr. 18): Relationships between the Homography and the Essential Matrix
- Lecture 6 (Apr. 20): Stratification in 2D: from Projective to Affine to Euclidean
- Lecture 7 (Apr. 25): Uncalibrated Epipolar Geometry
- Lecture 8 (Apr. 27): Projective Reconstruction
- Lecture 9 (May 2): Affine and Euclidean Reconstruction
- Lecture 10 (May 4): Camera Calibration
*Prof. Belongie out of town May 9 & 11 for ECCV 2006. Satya will run a two-part showing of the instructional DVD Computer Vision: Fact and Fiction.*- Lecture 11 (May 16): Optimal Pose and Structure
- Lecture 12 (May 18): Review Q&A Session
- Lecture 13 (May 23): Interest Point Detection
- Lecture 14 (May 25): Robust Feature Matching
- Lecture 15 (May 30): Feature Tracking and Optical Flow
- Lecture 16 (June 1): Orthographic Structure from Motion using Tracked Points
*no class June 6*- Lecture 17 (June 8): Review for Final Exam

Links:

- Announcements
- Scribe notes from CSE 252B Spring 2004
- eigshow.m, svdshow.m: Interactive exploration of eigenvalues and singular values (see also [pdf])
- Matlab information
- CSE 252B e-reserves
- Camera Calibration Toolbox for Matlab (Bouguet)
- Microsoft Camera Calibration Code (Zhang)
- Intel OpenCV
- CVonline
- The Computer Vision Home Page
- Related classes at UCSD: CSE 152, CSE 166, CSE 252A
- Slides/figures from Forsyth & Ponce, Hartley & Zisserman, Pollefeys

Required textbook:

An Invitation to 3D Vision: From Images to Geometric Models Ma, Soatto, Kosecka and Sastry Springer Verlag, 2003 ISBN 0-387-00893-4 (Errata as of October 2004) |

Multiple View Geometry in Computer Vision, 2nd edition by Hartley & ZissermanHandy Math reference:

Computer Vision -- A Modern Approach by Forsyth and Ponce

Introductory Techniques for 3-D Computer Vision by Trucco & Verri

The Geometry of Multiple Images by Faugeras, Luong, and Papadopoulo

Vision Science: Photons to Phenomenology by Stephen E. Palmer

MathWorld

*Most recently updated on May. 23, 2006 by Serge Belongie.*