Back to Search Document details
39th Meeting: Daejeon, KR, June 2025 2025-05-20 18:45
AHG7: Summary on the Tool Analysis in Earlier Proposals/Reports
Authors: Xinwei Li Google Xiang Li
Abstract
During the AHG7 call on Assessment Perspectives of Codec/Coding Tools between 39th and 40th JVET meeting, it was requested to further check the tool analysis in earlier proposals and reports, such as AHG5 and AHG16 activities during VVC development, AHG12, AHG9 activities during HEVC development. In this contribution, the activities are summarized.
JVET-AM0044 AHG7: Summary on the Tool Analysis in Earlier Proposals/Reports [X. Li (Google)]

During the AHG7 call on Assessment Perspectives of Codec/Coding Tools between 39th and 40th JVET meeting, it was requested to further check the tool analysis in earlier proposals and reports, such as AHG5 and AHG16 activities during VVC development, AHG12, AHG9 activities during HEVC development. In this contribution, the activities are summarized.

The contribution is a decent summary of methods used to assessment of tools complexity during HEVC and VVC standard development. It was commented that the VVC CfP H1002 and the proposal package description template JVET-H1003 also contained a lot of detailed aspects to be described about complexity of an algorithm and its tools, but there was no number criteria to be provided. It might be a good exercise to investigate how VVC could be quantitatively analysed by the criteria of the former CfP.

It was further commented that the previous CfP was too much focused on detailed building blocks of a classical hybrid codec. It appears more important to define more abstract criteria such as local memory, tables, dependencies at which granularity, capability for parallelization, etc., and some new criteria may need to be added e.g. for neural networks.

Decisions
It was further commented that the previous CfP was too much focused on detailed building blocks of a classical hybrid codec. It appears more important to define more abstract criteria such as local memory, tables, dependencies at which granularity, capability for parallelization, etc., and some new criteria may need to be added e.g. for neural networks.
Citation