Scholarlink

Versatile Video Coding (VVC/H.266), the latest video standard, introduces multiple new coding techniques. Among these, Multiple Transform Selection (MTS) aims to enhance transform coding efficiency in the encoder. However, MTS increases coding complexity. This paper presents a new lightweight AI-based method to streamline MTS in VVC using machine learning, specifically the Random Forest algorithm. Our method simplifies transform selection in the VVC module, replacing the original complex process. Experimental results using VVC reference software VTM-14 in random access configuration at QP 32 show that our algorithm reduces transform coding time by 39.44% and overall encoder time by 9%, with only a 1.6% increase in bitrate compared to the standard method.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Random Forest Based Fast MTS Algorithm for VVC Encoder

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sameh Samir,
  • Matthieu Saumard,
  • Taheni Damak,
  • Maher Jridi,
  • Mohamed Ali Ben Ayed

摘要

Versatile Video Coding (VVC/H.266), the latest video standard, introduces multiple new coding techniques. Among these, Multiple Transform Selection (MTS) aims to enhance transform coding efficiency in the encoder. However, MTS increases coding complexity. This paper presents a new lightweight AI-based method to streamline MTS in VVC using machine learning, specifically the Random Forest algorithm. Our method simplifies transform selection in the VVC module, replacing the original complex process. Experimental results using VVC reference software VTM-14 in random access configuration at QP 32 show that our algorithm reduces transform coding time by 39.44% and overall encoder time by 9%, with only a 1.6% increase in bitrate compared to the standard method.