Generating Video from Digital Photographs

Jonathan Foote, Matthew Cooper, John Adcock, and David Hilbert



We present methods for automatically turning digital photographs into a compelling video animation. Images are animated using dynamic effects like cropping, panning, zooming, rotation, and warping, synchronized with a user-selected soundtrack. Appropriate effects are automatically determined from features extracted from the images. For example, images with detected faces or strong symmetry are good candidates for a zoom effect. The system architecture flexibly supports other animation criteria such as pan direction from spatial frequency analysis. These dynamic effects are automatically synchronized to an audio soundtrack, resulting in a compelling video. We also consider lightweight output types such as Macromedia SWF and SMIL as well as more conventional video formats.

submitted for publication

(PDF, 649 kb)