“Multiple-View Geometry for Image-Based Modeling” by Kosecka, Vidal and Yu

  • ©Jana Kosecka, Rene Vidal, and Yizhou Yu

Entry Number: 42


    Multiple-View Geometry for Image-Based Modeling

Course Organizer(s):



    Basic knowledge of image formation and modeling, 2D and 3D geometry, rigid body transformations and linear algebra, from experience with image-based rendering, or from an introductory-level computer vision and graphics course. Unlike traditional treatment of this subject, this course will not require any knowledge of tensorial algebra or projective geometry.

    Key algorithms for building a 3D geometric model from 2D images. Image formation. Feature extraction, matching, and tracking. Two-view geometry. Uncalibrated geometry and reconstruction, geometry, and reconstruction from multiple views. Incorporation of scene knowledge. Step-by-step building of a 3D model. Image-based texture mapping and rendering from photographs.

    The state of the art in multiple-view geometry, including methods and algorithms for reconstructing 3D geometric models from video or photographs. This course is based on a novel approach to multiple-view geometry that requires only linear algebra, which makes image-based modeling techniques accessible to a larger audience.