Scholarlink

<p>As the latest video encoding standard, Versatile Video Coding (VVC) can achieve a 50%-bit rate savings for equal video quality compared to the previous generation High Efficiency Video Coding (HEVC) standard by introducing new technologies such as Multi-Type Tree (MTT) in the coding unit (CU) structure. But it also brings a huge computational burden, which limits its application. To address this problem, this paper proposes a Min-CU based fast CU partitioning algorithm for intra coding of VVC. Firstly, the complexity of feature computation is reduced by using the min-CU, i.e., 4 × 4 CU, as a minimum computational unit instead of the original luma pixel. Secondly, in order to express the texture information of CU comprehensively, in addition to the traditional overall texture features, this paper also proposes a novel local texture feature and uses both overall and local texture features to jointly characterize the texture intensity of CU. Finally, these texture features are inputted into the trained random forest model to determine whether CU needs to be segmented or not, and if so, a heuristic decision model is proposed to select the most appropriate one from five partitions. Experimental results show that the proposed algorithm is able to balance the reduction in computational complexity with the loss of coding quality and can reduce 53.38% encoding time with Bjøntegaard Delta Bit Rate (BDBR) only increasing by 1.63% over VTM-10.0.</p>

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

Min-CU based fast CU partitioning for VVC intra coding

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Xiangjie Song,
  • Yingbiao Yao,
  • Qing Liu,
  • Chenjie Du

摘要

As the latest video encoding standard, Versatile Video Coding (VVC) can achieve a 50%-bit rate savings for equal video quality compared to the previous generation High Efficiency Video Coding (HEVC) standard by introducing new technologies such as Multi-Type Tree (MTT) in the coding unit (CU) structure. But it also brings a huge computational burden, which limits its application. To address this problem, this paper proposes a Min-CU based fast CU partitioning algorithm for intra coding of VVC. Firstly, the complexity of feature computation is reduced by using the min-CU, i.e., 4 × 4 CU, as a minimum computational unit instead of the original luma pixel. Secondly, in order to express the texture information of CU comprehensively, in addition to the traditional overall texture features, this paper also proposes a novel local texture feature and uses both overall and local texture features to jointly characterize the texture intensity of CU. Finally, these texture features are inputted into the trained random forest model to determine whether CU needs to be segmented or not, and if so, a heuristic decision model is proposed to select the most appropriate one from five partitions. Experimental results show that the proposed algorithm is able to balance the reduction in computational complexity with the loss of coding quality and can reduce 53.38% encoding time with Bjøntegaard Delta Bit Rate (BDBR) only increasing by 1.63% over VTM-10.0.