JVET-AP0138 AHG9: On SEI messages presence [M. Damghanian, S. Harribey, M. Pettersson, R. Sjöberg (Ericsson)]

This contribution proposes new text to be added to VSEI v4 (errata) and VSEI TuC, to specify a preferred decoder action for the case that presence of an SEI message for the first picture of a CLVS (or CVS) is mandated when there is an SEI message of that type present for pictures other than the first picture in that CLVS (or CVS), and the SEI message for the first picture of the CLVS (or CVS) is not present (e.g., dropped or lost). In such a case, the proponents claim that it is not specified or suggested what action a decoder supporting the SEI message should take. The proponents state that a decoder supporting the SEI message could be configured for instance a) to act as if the SEI message for the first picture of the CVS were present, b) to ignore the SEI message, etc. The proponents claim that a decoder may benefit from a described preferred action to improve interoperability. Three options are proposed that can be applied to single SEI messages or a group of SEI messages for which an SEI message for the first picture of a CLVS, or CVS, is mandated.

It was commented that if a MANE were to drop a first SEI message in a CLVS, it would likely drop subsequent occurrences as well.

This proposal can address the network loss of an SEI message. We have not previously considered error concealment for SEI messages. The video coding standards also don’t recommend behaviour for lost pictures and slices.

It was considered dangerous to make such a change for existing standards.

It was suggested that a TR could address non-conforming bitstreams.