Contact
Director
Prof. Gerhard Fohler
Secretary
Stephanie Jung
Phone: +49 631 205 2086

Office Hours
Mon/Tue: 13:00-16:00
Wed-Fri: 08:00-11:00
You are here: RTS-Portal>>Publications

Publications

Search Publications

show/hide

Parallelizing MPEG-4 Video Decoding for Real-Time Processing on Multicore Systems


Category: Masters thesis
Author(s)Martin Brocksch
SourceTechnische Universität Kaiserslautern.
Date Dec 2009
Download:

Abstract

For the real-time decoding of high definition H.264 video streams a high compu- tational power is required, which imposes a serious challenge on current computer architectures. A convenient way to solve this problem is the use of multi-core systems. This thesis investigates the parallelization possibilities of H.264 video decoding and proposes a parallelization approach using data partitioning on mac- roblock layer. Therefore the data dependencies of the data partitions on mac- roblock layer are analysed using a dependecy graph representation and three mac- roblock scheduling algorithms are proposed. A parallel decoder implementation based on the libavcodec H.264 decoder which uses these algorithms is designed and evaluated by performing a number of measurements. The three algorithms are compared taking into account the number of processing cores and the effect of using dynamic or worst case dependencies for macroblock scheduling. The results show that macroblock level parallelism is a good encoder-independent method to speed up the reconstruction stage of H.264 video decoding. The overall speedup is however limited by the sequential execution of the entropy decoding stage.

Bibtex


Upcoming Events
Currently no events
Latest News:
03.07.17 10:13
Our paper titled "DREAMS Toolchain: Model-Driven Engineering of Mixed-Criticality...


17.06.17 12:08
20.06.2017, 11:45-13:15 @ 11-207 on "Microkernel-Based Systems" by Dr. Marcus Völp (SnT –...


16.06.17 21:40
Our extended abstract titled "DRAM-related Challenges in Task Scheduling with Timing...


All News