JVET-AG0184 AHG9: Text comment SEI message [P. de Lagrange, D. Doyen, E. François, F. Urban, C. Salmon-Legagneur (InterDigital)]

Chaired by J. Boyce.

This contribution proposes a new SEI message allowing to embed text comments in a video bitstream. Examples of use cases are annotations for parts of a video sequence, descriptive annotations for indexing purpose, user generated comments. In the proposed "text comment SEI message”, text comments are associated with ids to allow overlapping persistence (and potentially defining different types of comments).

Possible use cases include annotations, carrying technical data, user generated comments.

Has some similarity to AG0184 AI marking, but has different syntax for carrying text. AG0184 uses st(v), this contribution used b(8).

This contribution also allows overlapping text labelling, so persistence could differ for individual text comments.

H.263 Annex W had a feature for carrying text with different purposes.

Annotated regions SEI message also supports text labels, but limits to 255 characters. AR SEI also has an indication of what language is used. It was suggested that having a language tag could be useful.

Systems specs already have timed text feature, which includes presentation information for displaying the text. Would be useful to understand the systems timed text features better to understand any differences in capability with the contribution.

It would be possible to have a purpose id that allows text to be used for multiple purposes, potentially including copyright and AI marking. External means could potentially be used for signalling the purpose, for example based on the id. Another alternative is to use separate SEI messages for each purpose but have a consistent text signalling mechanism, perhaps using a common syntax function. There is not a shortage of SEI message payload type values.

Further study encouraged.