TitleuAVS2-Fast encoder for the 2nd generation IEEE 1857 video coding standard
AuthorsWang, Zhenyu
Wang, Ronggang
Fan, Kui
Sun, Huifang
Gao, Wen
AffiliationPeking Univ, Shenzhen Grad Sch, Shenzhen 518055, Guangdong, Peoples R China.
MERL, 201 Broadway, Cambridge, MA 02139 USA.
KeywordsIEEE 1857
AVS2
Video encoder
Fast algorithm
HEVC
CU SIZE DECISION
MOTION ESTIMATION ALGORITHM
INTER-MODE DECISION
HEVC
EFFICIENCY
SELECTION
PREDICTION
Issue Date2017
PublisherSIGNAL PROCESSING-IMAGE COMMUNICATION
CitationSIGNAL PROCESSING-IMAGE COMMUNICATION.2017,53,13-23.
AbstractThe recently issued second generation IEEE 1857 video coding standard (AVS2) provides a doubling in coding efficiency relative to previous standards such as H.264/AVC and AVS1. The coding performance improvement is mainly due to the adoption of advanced coding tools. However, these coding tools also cause a dramatic increase in computational complexity. A fast encoder design for AVS2 is strongly demanded to promote it being widely used in video compression practice. In this paper, we propose a fast encoder design for AVS2. In our design, we propose a new quantization step (QStep) dependent fast CU depth decision algorithm for inter frames, and a combination strategy to integrate fast algorithms in different coding Presets to get a good trade-off between coding speed and coding efficiency. In addition, instruction optimization, parallel coding and some coding flow optimizations are also used in our design. According to the proposed design, we developed the first fast AVS2 encoder, and named it as uAVS2. Experimental results show that uAVS2 is 56.58 times faster than the AVS2 reference software, and 19.32 times faster than the HEVC reference software, with negligible performance loss. uAVS2 is 2.7-6.6 times faster than the well-known fast HEVC encoder x265 under similar coding performance configurations. Under similar coding speed, the coding performance of uAVS2 is superior to x265 by 10-30% in BD-rate. uAVS2 with the highest speed Preset can encode 1080 P video in real-time on PC platform.
URIhttp://hdl.handle.net/20.500.11897/474378
ISSN0923-5965
DOI10.1016/j.image.2017.01.006
IndexedSCI(E)
Appears in Collections:深圳研究生院待认领

Files in This Work
There are no files associated with this item.

Web of Science®



Checked on Last Week

Scopus®



Checked on Current Time

百度学术™



Checked on Current Time

Google Scholar™





License: See PKU IR operational policies.