基于决策树的HEVC到VVC转码算法  

HEVC-to-VVC Transcoding Algorithm Based on Decision Tree

作　　者：许世扬 肖广[2] 滕国伟[1] 

机构地区：[1]上海大学通信与信息工程学院智慧考试联合实验室,上海200444 [2]上海市教育考试院,上海200433

出　　处：《工业控制计算机》2025年第1期66-67,128,共3页Industrial Control Computer

摘　　要：H.266/VVC优秀的编码效率促使人们将H.265/HEVC编码的视频内容转码到H.266/VVC标准。然而H.266/VVC新引入的四叉树嵌套多类型树(QTMT)编码单元(CU)划分技术显著增加了VVC编码可能的划分模式,并导致HEVC到VVC的转码的高复杂度。因此提出了基于深度重用和决策树的HEVC到VVC转码算法。首先通过HEVC的划分深度缩小了VVC编码单元划分的深度范围,随后使用决策树判断剩余深度上CU分裂的可能性,并对不太可能的划分深度执行简化测试。实验结果显示,和原版的转码算法相比,提出的算法平均降低了45.53%的转码复杂度,而BDBR仅增加了0.80%。The exceptional coding efficiency of H.266/VVC has motivated people to transcode H.265/HEVC-encoded video content to the H.266/VVC standard.However,the newly introduced Quadtree Nested Multi-Type Tree(QTMT) coding unit(CU) partitioning technique in H.266/VVC significantly increases the possible partitioning modes for VVC encoding,resulting in the high complexity of HEVC to VVC transcoding.Therefore,this paper proposes a HEVC to VVC transcoding algorithm based on depth reuse and decision tree.The depth range of VVC CU partitioning is first narrowed down by the partition depth of HEVC.Subsequently,a decision tree model is used to judge the likelihood of CU splitting at the remaining depths,and performing simplified tests for those partition depths that have low possibility to be reached.The experimental result shows that our proposed algorithm reduces the transcoding complexity by an average of 45.53%,with the BDBR increasing by only 0.80% compared to the original transcoding algorithm.

关 键 词：H.265/HEVC H.266/VVC 转码 

分 类 号：TN919.81[电子电信—通信与信息系统]