Abstract: The new History-based Motion Vector Prediction(HMVP) technology in H.266 cannot refer to the earlier motion information during the line-feeding update of the list, and the de-redundancy process of the HMVP candidate list when added into the Merge list is cumbersome and time-consuming.To address the problem, this paper proposes an improved candidate decision making algorithm for the merge mode in H.266.The algorithm temporarily stores a HMVP candidate list to solve the defects of the original HMVP list when it is cleared in line breaks, so as to better predict the motion information.Then the number of original spatial-temporal candidates in the Merge candidate list is used to determine the number of redundant HMVP candidates, so the de-redundancy process can be simplified.The experimental results show that compared with the standard algorithm of H.266, the improved algorithm saves 0.46% of the code rate on the luminance component of the test image sequence in the random access configuration.It reduces the total coding time by 4.09% on average, and the inter-frame prediction time by 10.87% on average.The improved algorithm improves coding performance while reducing coding complexity.

Key words: Versatile Video Coding(VVC), inter-frame prediction, History-based Motion Vector Prediction(HMVP), merge mode, random access

摘要： 针对H.266中基于历史块的运动矢量预测（HMVP）技术在换行更新列表过程中无法参考之前运动信息，以及HMVP候选列表在加入Merge列表过程中存在繁琐的去冗余过程的问题，提出一种改进的H.266合并模式候选决策算法。通过临时存储一个HMVP候选列表来解决原有HMVP列表在换行清空时存在的无法参考之前运动信息的不足，从而更好地预测运动信息。通过Merge候选列表中原有的空时域候选数量来确定需要进行去冗余的HMVP候选数量，以简化去冗余过程。实验结果表明，与H.266标准算法相比，该算法在随机接入的配置下，测试图像序列的亮度分量能够平均节省0.46%的码率，编码总时间平均缩短4.09%，帧间预测时间平均缩短10.87%，能够提高编码性能同时降低编码复杂度。

关键词: 通用视频编码, 帧间预测, 基于历史块的运动矢量预测, 合并模式, 随机接入

CLC Number: 

• TP391