NORMA eResearch @NCI Library

Many-Core HEVC Encoding Based on Wavefront Parallel Processing and GPU-accelerated Motion Estimation

Radicke, Stefan, Hahn, Jens-Uwe, Wang, Qi and Grecos, Christos (2015) Many-Core HEVC Encoding Based on Wavefront Parallel Processing and GPU-accelerated Motion Estimation. In: E-Business and Telecommunications. ICETE 2014. Communications in Computer and Information Science (554). Springer, Cham, pp. 393-417. ISBN 9783319259154

Full text not available from this repository.
Official URL: http://dx.doi.org/10.1007/978-3-319-25915-4_21

Abstract

The High Efficiency Video Coding (HEVC) standard provides an outstanding compression performance and is thus ideally suited for Ultra High Definition (UHD) content. However, the complexity of the encoder is substantial and therefore highly optimized implementations are required to achieve reasonable speeds. For this purpose, high-level parallelization mechanisms like Wavefront Parallel Processing (WPP), can be used to leverage modern multi-core hardware. In this work, the WPP mechanism is theoretically analyzed and a non-intrusive implementation of it based on the reference test model HM-13.0 is presented. Furthermore, a novel extension for heterogeneous computing platforms called Heterogeneous WPP (HWPP) is proposed which largely increases the achievable speedups. To demonstrate the power of HWPP, a Graphics Processing Unit (GPU) accelerated Motion Estimation (ME) algorithm is integrated. Based on a large amount of experimental data, it is shown that the speedups achieved with WPP and HWPP reach up to 8.9 and 17.9 times, respectively.

Item Type: Book Section
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Divisions: School of Computing > Staff Research and Publications
Depositing User: Caoimhe Ní Mhaicín
Date Deposited: 28 Feb 2019 14:40
Last Modified: 28 Feb 2019 14:40
URI: https://norma.ncirl.ie/id/eprint/3615

Actions (login required)

View Item View Item