Computational & Technology Resources
an online resource for computational,
engineering & technology publications
PROCEEDINGS OF THE FIFTH INTERNATIONAL CONFERENCE ON PARALLEL, DISTRIBUTED, GRID AND CLOUD COMPUTING FOR ENGINEERING
Edited by: P. Iványi, B.H.V. Topping and G. Várady
Heterogeneous Parallel Approaches for HEVC Encoder
H. Migallón, O. López-Granado, V. Galiano, P. Piņol, M. Martinez-Rach and M.P. Malumbres
Miguel Hernández University, Elche, Spain
H. Migallón, O. López-Granado, V. Galiano, P. Piñol, M. Martinez-Rach, M.P. Malumbres, "Heterogeneous Parallel Approaches for HEVC Encoder", in P. Iványi, B.H.V. Topping, G. Várady, (Editors), "Proceedings of the Fifth International Conference on Parallel, Distributed, Grid and Cloud Computing for Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 41, 2017. doi:10.4203/ccp.111.41
Keywords: parallel algorithm, video coding, HEVC, hierarchical parallelism, multicore, GOP, slice partitioning, distributed memory, shared memory.
Several attempts have been made in order to speed-up the total computational time required by the HEVC standard to encode a video sequence. In this paper, we propose a hierarchical parallelization approach in the HEVC encoder based on GOPs and slices so as to better exploit the hardware resources available in the parallel architecture and to significantly reduce the total encoding time of a video sequence. The main idea in this hierarchical approach is to divide the video sequence in GOPs or group of GOPs that will be assigned to each computing node. Inside each node, every frame or picture that belongs to a GOP will be split in as many number of slices as the number of computing processes in that node. In this manner we will be able to take advantage of both distributed and shared memory parallelism. The results show that speed-ups of up to 68x (using 10 nodes and 11 processes per node) can be obtained when combining both GOPs and slices techniques for high resolution video sequences.
purchase the full-text of this paper (price £22)