JVET-AC0077 AHG9/AHG15: On bitstreams that are potentially suboptimal for user viewing [M. M. Hannuksela, F. Cricri, H. Zhang (Nokia)]
The contribution reviews some earlier contributions on preprocessing and encoding methods targeted at machine analysis and concludes that spatial and temporal quality fluctuations resulting from such methods may result into decoded video that looks subjectively unpleasant.
The contribution proposes a suboptimal user viewing indication SEI message, which indicates that the video resulting by decoding the current CLVS is not optimized for user viewing and may therefore look incoherent. The proposed SEI message can be used to avoid accidental displaying of machine-targeted bitstreams.
It was pointed out that at systems layer in file format specification there is an indicator that a video is not suitable for viewing (entitled as “restricted video”)
In the discussion, specifically in the context of VCM, it was suggested to clarify with WG 4 how metadata for machine consumption would be conveyed, which will likely be needed also in the case when die video stream itself is a legacy stream.
The contributors explain that the usage of the proposed SEI message is not specific to the VCM application. The message also contains indicators which convey if only some temporal sublayers are suboptimal. It was asked if in such concept it might also be beneficial to indicate if some regions are or are not useful. It was noted that the annotated regions SEI might fulfill such purpose (see also discussion in BoG JVET-AC0324).
It was further noted that concept of VUI indication exists to indicate that a bitstream has some special meaning. It might be more consistent to align the intended purpose with that.
Further study was recommended.
JVET-AC0077 AHG9/AHG15: On bitstreams that are potentially suboptimal for user viewing [M. M. Hannuksela, F. Cricri, H. Zhang (Nokia)]
See notes in section 4.14.