Cost-efficient HEVC-based quadtree splitting (HEQUS) for VVC Video Transcoding

作者:

Highlights:

• HEQUS is a fast HEVC-to-VVC transcoding algorithm based on machine learning.

• A Naïve-Bayes model assists the partitioning of the first level of the quadtree.

• The remaining quadtree levels are based on the original HEVC partitioning.

• HEQUS achieves 44% encoding time reduction at the expense of only 3% BD-rate.

• Alternatively, Fast HEQUS achieves 60% time savings by replicating the HEVC quadtree.

摘要

•HEQUS is a fast HEVC-to-VVC transcoding algorithm based on machine learning.•A Naïve-Bayes model assists the partitioning of the first level of the quadtree.•The remaining quadtree levels are based on the original HEVC partitioning.•HEQUS achieves 44% encoding time reduction at the expense of only 3% BD-rate.•Alternatively, Fast HEQUS achieves 60% time savings by replicating the HEVC quadtree.

论文关键词:HEVC,H.265,VVC,Transcoding,MTT

论文评审过程:Received 15 February 2020, Revised 2 December 2020, Accepted 4 February 2021, Available online 12 February 2021, Version of Record 17 February 2021.

论文官网地址:https://doi.org/10.1016/j.image.2021.116199