JVET-AP0050 AHG17: Proposal of target bitrates for the HDR and UHD Gaming sequences for the upcoming CfP [C. Lehmann, A. Wieckowski, B. Bross, J. Pfaff (Fraunhofer HHI), Z. Li, J. Sauer, T. Solovyev, E. Alshina (Huawei)]

Sequences included in JVET-AP0279.

For the earlier versions of this document no need for presentation was foreseen, but -v6 had substantial changes including a suggestion to modify the CfP. This was presented and discussed on Wednesday 29 April 1730-1805.

Abstract:

In this proposal, the proponents propose and update a set of UHD SDR and HD HDR gaming sequences to be evaluated as additional candidates for the upcoming CfP. The contribution comprises three distinct sets of synthetic sequences. The Wukong sequences, which have already been evaluated and recommended as candidates, will be presented as an update with rate-matched results for the latest VTM. The Carla sequences have been adapted based on feedback from the subjective viewings in AHG17. Additionally, three new HD HDR gaming sequences, already part of the Gaming CTC, are introduced for evaluation for the next-generation video coding standard. After subsequent viewing of the sequences at the 42nd meeting, it was agreed to add Wukung2 and Carla5 into the CfP draft and do not evaluate the new HD HDR gaming sequences subjectively. HDR gaming, especially with the ground truth auxiliary metadata, is a relevant use case that should be covered and evaluated in some way at the CfP as well. Hence, it is proposed to evaluate this HD HDR gaming category only objectively as for the submitted R5 rate points. Furthermore, we consider streaming of gaming content in a RA configuration a relevant use case and propose to also evaluate this category objectively as well. A proposed modified draft CfP is attached to this document.

From the discussion:

It was commented that the main problem in the gaming category is the missing variety of content (as concluded in the meeting in Aachen), which is an important aspect for future standard CTC, and might also be looked at in the selection of the supplementary set (not necessarily from the set of sequences that have been known for a long time), where also RA configurations of gaming sequences could be conducted. Adding more sequences of Wukong and Carla might not be beneficial in this context.

It was commented that the aspect of using metadata obtained from gaming sequence rendering is still under investigation in AHG15. Auxiliary data are only available only for some sequences, investigation about potentially generating them artificially by estimation is ongoing (see section 4.14), and it was already concluded in the last meeting that it would be premature to attach them to the CfP sequence set, even if they were available for some sequences. It is understood that this could become an important topic during standard development. Also, supplemental information about benefits by using metadata of gaming sequences would be highly welcome in the context of a proposal.

No action at this stage.