Back to Search Document details
40th Meeting: Geneva, CH, October 2025 2025-10-12 16:33
AHG18: CfE response in additional functionality on ultra-low latency and packet loss resilience
Authors: Sergey Ikonin Huawei Viacheslav Khamidullin Huawei I. Gribushin Huawei M. Sychev Huawei K. Malyshev Huawei Akhsarbek Dzugaev Huawei A. Bovsha Huawei Xiang Ma Huawei Elena Alshina Huawei
Abstract
This contribution describes the coding technology beyond VVC that can be used for ultra-low latency and packet loss resilient decoding scenarios. The described technology re-uses concept of data partitioning of H.263/4, which separates prediction and residual information into different sub-streams. Additionally, entropy coding of residual coefficients is performed using multiple independent entropy coded chunks. A lost aware NN-based filter is further applied to reduce subjective impact of lost residuals and propagation if these errors via inter frame prediction. The evaluation methodology following discussion and contributions of AHG18 is also proposed. According to this methodology the described solution provides average 2.8 dB improvement on top of scalable VTM anchor in broadcast scenario, and freeze rate reduction from 14.7% to 0.4% over single layer VTM anchor in unicast scenario.
Meeting Notes
JVET-AN0080 AHG18: CfE response in additional functionality on ultra-low latency and packet loss resilience [S. Ikonin, V. Khamidullin, I. Gribushin, M. Sychev, K. Malyshev, A. Dzugaev, A. Bovsha, X. Ma, E. Alshina (Huawei)]

This contribution describes the coding technology beyond VVC that can be used for ultra-low latency and packet loss resilient decoding scenarios. The described technology re-uses concept of data partitioning of H.263/4, which separates prediction and residual information into different sub-streams. Additionally, entropy coding of residual coefficients is performed using multiple independent entropy coded chunks. A lost aware NN-based filter is further applied to reduce subjective impact of lost residuals and propagation if these errors via inter frame prediction. The evaluation methodology following discussion and contributions of AHG18 is also proposed. According to this methodology the described solution provides average 2.8 dB improvement on top of scalable VTM anchor in broadcast scenario, and freeze rate reduction from 14.7% to 0.4% over single layer VTM anchor in unicast scenario.

NN filter not yet implemented in SADL.

Overhead is counted in NAL units.

It was commented that it might be appropriate to also use channel conditions of other networks.

Data partitioning and entropy coding would need to be normative. NN-based filter might not need to be defined normatively, as it does not change the output in the case of uncorrupted bitstream.

It was commented that a key question would be if all decoders would need to be burdened by tools that are only relevant for some applications.

In case of unicast, the feedback to encoder is at frame level, not packet level.

What is the granularity of data partitioning? Block (CU) level. However, partitioned information from multiple CUs is transported in separate NAL units. The positions of lost CUs are assumed to be known for further decoding operations.

The opinion is expressed that the testing method needs improvement.

JVET-AN0080 AHG18: CfE response in additional functionality on ultra-low latency and packet loss resilience [S. Ikonin, V. Khamidullin, I. Gribushin, M. Sychev, K. Malyshev, A. Dzugaev, A. Bovsha, X. Ma, E. Alshina (Huawei)]

See section 4.16.2.

Subsequent documents were reviewed on Thursday 9 Oct.

Cited:
JVET-AO0048
ARCHIVE :
BroadcastAnchor.csv BroadcastTest.csv Demo.Broadcast.Beatriz.Lat50.mp4 Demo.Unicast.OfficeWalkAtWall.Lat24.mp4 JVET-AN0080-v1.pptx JVET-AN0080-v1.xlsm JVET-AN0080-v2-clean.docx JVET-AN0080-v2.docx JVET-AN0080_APS_high_prioiry-v1.xlsm UnicastAnchor.csv UnicastTest.csv
Decisions
The opinion is expressed that the testing method needs improvement.
Subsequent documents were reviewed on Thursday 9 Oct.
Citation