Back to Search Document details
14th Meeting: Geneva, March 2019 2019-03-19 11:21
JVET AHG report: Progressive intra refresh (AHG14)
Abstract
This document summarizes activities of AhG on Progressive intra refresh between the 13th and the 14th JVET meetings.
JVET-N0014 JVET AHG report: Progressive intra refresh (AHG14) [J.-M. Thiesse, A. Duenas, K. Kazui, R. Sjöberg, A. Tourapis]

This document summarizes activities of AhG on Progressive intra refresh between the 13th and the 14th JVET meetings.

There were several mail exchanges on the main JVET email reflector (jvet@lists.rwth-aachen.de) with an [AHG14] indication on message headers. Three topics had been discussed on these mails:

  • First, the relevance of intra refresh for low delay has been discussed. Thanks to the early sharing of a latency study by Huawei, the appropriateness of intra refresh was demonstrated without or together with DU.
  • Second, there were emails regarding the opportunity of having normative changes to support intra refresh. One expert indicates that the potential additional checks for the decoder may be complex to support and this should be justified by a very high performance benefit threshold. Another expert indicated a use-case where having a smaller latency could be relevant as well as having a granularity to balance between latency and coding efficiency.

Another expert said that practical implementation using intra refresh are complex for handling of entertainment/broadcasting applications. In particular, scene changes and commercial insertion were mentioned. Also, this expert and other expert commented on the complexity due to camera motions which are not in the same direction than the intra refresh scanning.

  • Third, there was a discussion on modifying the test condition which leads to the following conditions being agreed and proposed:
  • Basis = Low-Delay B Main10 configuration.
  • Do not use a pyramidal B structure (=> GOPSize = 1) in order to have all pictures with a somewhat similar bits sizes.
  • Use a QP offset of +1 between the first I and the first B picture (the B picture having a higher QP than the I picture).
  • Set the intra refresh period equal to 1 second, meaning that the stream shall be refreshed after a cycle of 1 second (can be less).
  • PSNR and bitrates should be computed without the first I picture + the first cycle minus 1 (meaning that the first 30 pictures are not considered in 30Hz, 60 in 60Hz…)

An updated version of the encoder only implementation has been shared on the AHG 14 Gitlab (https://vcgit.hhi.fraunhofer.de/jvet-l-ahg-14/VVCSoftware_VTM/tree/AHG14-IntraRefresh_EncOnly).

The following contributions were identified for the AHG.

  • Contribution on general aspects
    • JVET-N0114, “AHG14: A delay analysis for IRAP and GDR”, Hendry, Y.-K. Wang, M. Sychev (Huawei)
  • Contributions on normative intra refresh handling including software
    • JVET-N0080, “AHG14: Revised software for ultra low-latency encoding”, K. Kazui (Fujitsu)
    • JVET-N0391, “AHG14: Updates on intra refresh proposal”, J.-M. Thiesse, D. Gommelet, D. Nicholson (VITEC)
  • Contributions on normative intra refresh handling
    • JVET-N0115, “AHG14/AHG17: GDR - gradual decoding refresh”, Hendry, Y.-K. Wang, J. Chen, S. Hong (Huawei)
    • JVET-N0116, “AHG14/AHG17: DDR - distributed decoding refresh”, Y.-K. Wang, M. Sychev, Hendry (Huawei)
    • JVET-N0275, “AHG14/AHG8/AHG17: Sections for intra refresh and inloop filter disabling”, J. Boyce, R. Lei, L. Xu
    • JVET-N0310, “AHG14: On gradual decoding refresh”, M. M. Hannuksela, K. Kammachi-Sreedhar, A. Aminlou (Nokia)
  • Contributions on clean random access in case of intra refresh
    • JVET-N0101, “AHG14/ AHG17: On Gradual and Clean Random Access”, S. Deshpande (Sharp)
    • JVET-N0495, “AHG14: Recovery point indication NAL unit”, R. Sjöberg, M. Pettersson, M. Damghanian (Ericsson)

The AhG recommended:

  • To review all related contributions.
  • To integrate the encoder-only modifications into the next VTM software, update the anchor appropriately, and to continue investigating the benefits of normative progressive intra refresh approaches.
  • To consider updating the common test conditions for integrating the intra refresh scenario.
Decisions
To consider updating the common test conditions for integrating the intra refresh scenario.
Citation