跳到主要导航 跳到搜索 跳到主要内容

Prediction with multi-cross component

  • Junru Li
  • , Li Zhang
  • , Kai Zhang
  • , Hongbin Liu
  • , Meng Wang
  • , Shiqi Wang
  • , Siwei Ma
  • , Wen Gao

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

Cross-component prediction reveals great benefits in promoting the coding efficiency by removing the inter-channel redundancies between luma and chroma components. This paper proposes a novel prediction with a multi-cross component (PMC) mode for AVS3. With the PMC mode, the Cr coding block can be predicted through an efficient linear combination of the reconstructed coding blocks regarding Y and Cb components. Moreover, to release the computational burden, the linear model parameters in PMC are acquired by inheritance and reorganization. In this way, the inter-channel redundancies can be further removed, leading to reductions of the residual energies and improvements of coding performance. Experimental results on AVS3 test model HPM-5.0 verify the effectiveness of the proposed method where the PMC achieves on average 0.27%, 0.56%, and 1.10% BD-Rate savings for Y, Cb, and Cr components under all intra configuration. Moreover, the PMC brings 0.12%, 0.44% and 1.00% BD-Rate gains for Y, Cb, and Cr components under random access configuration. The variations of encoding and decoding time are negligible. Owing to the substantial gains regarding the compression efficiency, and low operational complexity, the PMC has been adopted to AVS3 standard.

源语言英语
主期刊名2020 IEEE International Conference on Multimedia and Expo Workshops, ICMEW 2020
出版商Institute of Electrical and Electronics Engineers Inc.
ISBN(电子版)9781728114859
DOI
出版状态已出版 - 7月 2020
已对外发布
活动2020 IEEE International Conference on Multimedia and Expo Workshops, ICMEW 2020 - London, 英国
期限: 6 7月 202010 7月 2020

出版系列

姓名2020 IEEE International Conference on Multimedia and Expo Workshops, ICMEW 2020

会议

会议2020 IEEE International Conference on Multimedia and Expo Workshops, ICMEW 2020
国家/地区英国
London
时期6/07/2010/07/20

指纹

探究 'Prediction with multi-cross component' 的科研主题。它们共同构成独一无二的指纹。

引用此