Multicore processor systems are leading the microprocessor industry today. This has placed more pressure upon programmers to write parallel programs that wisely balance load among cores on the same system and optimize performance. On the other hand, the video-form of data is getting more and more importance. Video is the fuel for many contemporary Internet applications like YouTube. Video is the most storage and bandwidth-hungry type of data, especially in the context of new video applications like HDTV and IPTV. Exploiting video compression manifests itself in real-time applications. In light of this, it is important to bring to practice parallelized video codecs programmed to run on multicore systems. In this paper, the authors concentrate on one aspect of the MPEG-4 video codec, the global motion estimation and compensation. They present a parallel implementation for MPEG-4 global motion estimation and compensation on multicore processors and provide a detailed performance evaluation under various scenarios.
Fast gradient methods based on global motion estimation for video compression