加权
计算机科学
编码(社会科学)
滤波器(信号处理)
运动补偿
计算机视觉
运动估计
人工智能
算法
数学
统计
医学
放射科
作者
Jack Enhorn,Christopher Hollmann,Rickard Sjöberg,Kenneth Andersson,Per Wennersten
标识
DOI:10.1109/iscas48785.2022.9937263
摘要
This paper presents improvements to the original motion compensated temporal pre-filter implemented in the VVC reference model VTM and the HEVC reference model HM. The improvements involve motion estimation modifications, increasing the number of neighboring pictures to use, and filtering strength weighting depending on block error metrics. Results are presented on VTM-14.0. Tested content includes JVET common conditions test set, sequences with introduced fading and scene cuts, and additional 4:2:2 and 4:4:4 chroma format test sequences.
科研通智能强力驱动
Strongly Powered by AbleSci AI