UCLA Extension

3D Modeling and Reconstruction from Video

A 3-Day Short Course

This course presents methods and algorithms to construct 3D models of scenes from video or photographs. Applications of this technology range from special effects (scene capture, camera motion capture, virtual insertion, image-based modeling and rendering) to security (video surveillance, photogrammetry, visual recognition); autonomous robotics; medical imaging; embedded sensor networks; and virtual reality.

Upon completing the course, participants should possess the tools necessary to capture imaging data (still or video) and process it to reconstruct the 3D shape of objects in a scene; to simulate 3D (rigid) motion; and to calibrate cameras.

The course features homework exercises and instruction is heavily based on the course text.

Designed for

  • Special effects editors, programmers, engineers
  • Studio engineers
  • Developers of computer graphics tools
  • Defense engineers
  • Robotics and automation engineers and practitioners
  • Visual effects engineers
  • Professional photographers and post-production editors
  • Surveillance engineers
  • Autonomous systems engineers
  • Engineers of automatically guided systems

Course Materials

The text, An Invitation to 3D Vision: From Images to Geometric Models, Y. Ma, S. Soatto, J. Kosecka, and S. Sastry (Springer Verlag, in press); sample software; and lecture notes are distributed on the first day of the course. The notes and software are for participants only and are not for sale.

Coordinator and Lecturer

Stefano Soatto, PhD, Associate Professor, Department of Computer Science, Henry Samueli School of Engineering and Applied Science, UCLA. Dr. Soatto joined the UCLA faculty in 2000 after serving as assistant/associate professor in electrical and biomedical engineering at Washington University, and as research associate in applied sciences at Harvard University. Dr. Soatto’s current research interests include computer vision and Nonlinear Estimation and Control Theory. In particular, he is interested in ways that computers can use sensory information (e.g., vision, sound, touch) to interact with humans and the environment. Dr. Soatto is the recipient of the David Marr Prize (with Y. Ma, J. Kosecka, and S. Sastry of UC Berkeley) for work on Euclidean reconstruction and reprojection up to subgroups. He also received the Siemens Prize with the Outstanding Paper Award from the IEEE Computer Society for his work on optimal structure from motion (with R. Brockett of Harvard), the National Science Foundation Career Award, and the Okawa Fellowship grant.

Course Program

Introduction

Fundamentals of Euclidean Geometry

  • Norms, inner products
  • Rigid motions

Light, Surfaces, Image Formation

  • Projections
  • Radiance, irradiance, reflectance
  • Lenses

Feature Detection and Tracking

  • Image feature detection
  • Templates and cross-correlation
  • Tracking
  • Robust correspondence

Epipolar Geometry and Stereo Reconstruction

  • Essential matrices
  • Epipolar constraints
  • 3D reconstruction from two views

Camera Calibration—With and Without a Rig

  • Calibration with a known object
  • Autocalibration

Fundamentals of Kalman Filtering

  • Linear Gaussian models
  • Kalman filter
  • Extended Kalman filter

Dynamic Vision; Motion Estimation

  • Dynamic motion estimation
  • Structure estimation
  • Occlusions
  • Outlier rejection

For more information contact the Short Course Program Office:
shortcourses@uclaextension.edu | (310) 825-3344 | fax (310) 206-2815

image_print