Simplified Rate-Distortion Optimized Quantization for HEVC

Motong Xu, Thuong Nguyen Canh, Byeungwoo Jeon

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

7 Scopus citations

Abstract

Rate-distortion optimized quantization (RDOQ) is shown to provide significant coding gain in HEVC but it also suffers from huge computational complexity. In this paper, we simplify the RDOQ tool in HEVC by enforcing more zero levels at high frequency locations without sacrificing coding efficiency much. First, we analyze the properties of each quantized level at different frequency locations. Subsequently, we set the quantized level of value 1 to 0 at specified high frequency locations to save the RDOQ processing time. In addition, a smaller round offset for quantization is adaptively used to minimize coding loss at high frequency locations. Our scheme is shown to reduce total quantization time by an average of 15.6% with only 0.01% loss in BDBR under the all intra coding configuration, and 13.17% of average time saving with 0.07% BDBR loss under the Random Access configuration.

Original languageEnglish
Title of host publication2018 IEEE International Symposium on Broadband Multimedia Systems and Broadcasting, BMSB 2018
PublisherIEEE Computer Society
ISBN (Print)9781538647295
DOIs
StatePublished - 14 Aug 2018
Externally publishedYes
Event13th IEEE International Symposium on Broadband Multimedia Systems and Broadcasting, BMSB 2018 - Valencia, Spain
Duration: 6 Jun 20188 Jun 2018

Publication series

NameIEEE International Symposium on Broadband Multimedia Systems and Broadcasting, BMSB
Volume2018-June
ISSN (Print)2155-5044
ISSN (Electronic)2155-5052

Conference

Conference13th IEEE International Symposium on Broadband Multimedia Systems and Broadcasting, BMSB 2018
Country/TerritorySpain
CityValencia
Period6/06/188/06/18

Keywords

  • HEVC
  • rate-distortion optimized quantization (RDOQ)
  • simplified RDOQ

Fingerprint

Dive into the research topics of 'Simplified Rate-Distortion Optimized Quantization for HEVC'. Together they form a unique fingerprint.

Cite this