Yasir Sinada

Olivier Steiger

18-796

For our multimedia communications project, we would like to treat some aspects of the video object planes (VOP) used in the MPEG-4 compression standard. Our main focus will be on different kinds of shape coding. We will then apply these algorithms to different types of data and try to establish some rules for their usage. However, we will not be concerned by shape recognition issues and therefor assume that the shape is given (chroma key).

The MPEG-4 conform technique that we will try to implement is gray-scale coding, where a plane is represented by a certain luminosity. This information will then be compressed using lossless as well as lossy algorithms, in order to find tradeoffs for different applications. Also, although it is not part of the MPEG-4 standard, we would like to compare the efficiency of gray-scale coding with a vector shape description. Once the shape is coded, our concern will be to implement its motion using motion estimation.

Our last step should then be to re-paste the video object into its shape, making use of some pre-existing codecs.

The work plan will be as follow:

1. By the midterm, all the shape intracoding techniques should be implemented and compared using standard test sequences.
   We will also have the necessary knowledge about motion estimation.
2. By the end of the semester, the whole motion-shape coding must be complete and we will hopefully provide some compression for different applications. If possible, we will provide a codec for a complete video-object sequence (talking head, tennis player, …).

Doing this work, we hope to acquire some detailed comprehension about a recent research subject and to try out a topic that is not yet "common world".