CSE 590SS, Winter 2001
M/W 3-4:20pm, MGH 228
Instructors: Steven Seitz and Richard Szeliski
Realistic image synthesis is a central goal of computer graphics. Movies like Jurassic Park or Star Wars demonstrate thrilling possibilities - graphical models that look and move so realistically that they integrate seamlessly with live action footage. Yet, creating such effects currently requires great artistry and painstaking manual labor. Motivated by these difficulties, many in the computer graphics community are turning to the field of computer vision as a means for capturing the real world directly from photographs and video. Beyond providing a rich source of input for computer graphics, computer vision has the potential to impact computer graphics at a variety of levels. For example, a camera-equipped PC could interpret your gestures and motions directly, without the need for a mouse or 3D-input device. Artists and media producers could benefit from vision-assisted tools that simplify editing images and video. Rendering architectures could take advantage of image-based representations to render complex scenes more efficiently.
In this course we will survey many of the computer vision techniques that have applications to the field of computer graphics research and production. The topics covered include image warping, matte extraction, motion estimation, mosaics, camera calibration, match move, shape recovery, texture analysis, and reflectance modeling. No prior background in computer vision is assumed. The fundamental concepts and mathematics that underlie these approaches will be covered in addition to the algorithms themselves.
Prerequisites: a prior course in computer graphics OR computer vision. The course is open to graduate students and advanced undergraduates (with instructor consent).
Last modified 03/07/2001