Motion Prediction in the GTV 2.0 Video Coder

by
Matt Renzelmann

In an uncompressed form, digital video is among the most storage- and bandwidth-intensive applications of modern computers. The high degree of redundancy among consecutive frames of such video, however, causes most of these resources to be wasted. Motion prediction is one technique for exploiting this redundancy, thereby allowing the video to be compressed more effectively. We will discuss the principal ideas of motion prediction, with an emphasis on the method used in our video coder, Group Testing for Video. In particular, we will discuss the effect of interpolation, variable block size prediction, and multiple reference frames. We will make comparisons between our method and those used in standards such as H.263 and H.264.

Advised by Richard Ladner

CSE 403
Wednesday
January 12, 2005
3:30 - 4:20 pm