JVET-R0041 AHG8/AHG9: On picture types and related constraints [Y.-K. Wang (Bytedance), R. Sjöberg, M. Pettersson, M. Damghanian, Z. Zhang, J. Enhorn, R. Yu (Ericsson)]
Discussed in AHG Session 1.9 (chaired by GJS).
This contribution proposes some changes related to definitions of the terms "associated IRAP picture", "associated GDR picture", and "trailing picture" and the constraints regarding different types of pictures and their relationships in terms of decoding order, output order, and prediction relationship.
The proposed changes are summarized as follows:
- The definition of associated GDR picture is added and the definition of associated IRAP picture is updated, such that each picture of a layer, except the first picture in the layer in the bitstream, is specified to be associated with the previous IRAP or GDR picture of the same layer in decoding order, whichever is closer.
- The definition of trailing picture is updated, such that a trailing picture may also be associated with a GDR picture.
- The following existing constraints are updated such that they only impose restrictions to pictures within the same layer:
- On the output order of pictures preceding an IRAP picture in decoding order
- On the decoding order of pictures associated with an IRAP picture and some non-leading pictures
- On RPLs for a CRA picture
- Constraints for an STSA picture, in terms of relative decoding order, output order, and prediction relationship with the associated IRAP picture and the same-layer pictures in the preceding and succeeding AUs, are specified, similarly as a trailing picture.
- Similar constraints for IRAP pictures and the same-layer pictures in the preceding and succeeding AUs in terms of relative decoding order, output order, and prediction relationship are specified for GDR pictures.
It was suggested to explicitly distinguish a “single-layer bitstream” from a “multi-layer bitstream”. As proposed, there may not be a need for such a distinction.
Some action is needed in this area; this is basically filling in gaps and proposing bug fixes for expression of the existing technical intent; no new functionalities are introduced.
Contribution JVET-R0226 is related.
It was commented that there may need to be some further cleanup regarding mixed NAL unit types; there are other contributions on that issue.
It was commented that it may be desirable to find a way to simplify / unify the discussions of GDR and IRAP if feasible (esp. if the recovery POC count is zero for the GDR).
It was noted that GDR characteristics are less restricted than IRAP, e.g. in regard to IRAP. The proponent said they had not intended to add constraints to GDR pictures, beyond what is necessary.
It was noted that various definitions depend on NAL unit types, and this could make the definitions difficult to comprehend outside the context of the document (a characteristic preferred by ISO editing guidelines).
AHG Recommendation: Adopt. (Further offline review is encouraged, and interacting aspects with other contributions remain in need of consideration.). See future details in the notes for JVET-R0226.