Matching pursuit for compression and application to motion compensated video coding

In many hybrid video compression algorithms there is a clear distinction between motion estimation and compensation (MEMC) on the one hand, and transform coding of the residue on the other hand. By translating MEMC into the framework of matching pursuit (MP), we will show that unification of both steps is possible. Moreover, adapting the MP approach to the small size dictionaries involved in MEMC, it is shown that unification may in principle lead to a better performance of hybrid coding schemes. We close with a discussion on the practical obstacles of the technique, and how to avoid them.<<ETX>>