JVET-AN0208 AHG8/AHG9: Depth-aware optimization for Encoder optimization information SEI message [G. Teniou, S. Wenger, A. Hinds, J. Ricard (Tencent)]

This document is a revision of the proposal for a new eoi_type associated to depth-aware encoding optimization. The comments made during the previous meeting are addressed and an updated version of the depth-aware information is proposed together with use case examples to illustrate the added value of the depth-aware encoder optimization information.

It was suggested that the EOI message doesn’t currently have a foreground/background optimization indication, although one of the eoi_object_based_idc value indicates areas outside detected objects have higher QP.

If the depth map is not provided in the bitstream, it is more difficult for the receiver to use this information.

It as suggested to provide a depth threshold to differentiate between foreground and background.

It is noted that source content may have focus at different depths.

It was suggested to include depth information using the eoi_object_based_idc, which is addressed in section 6 of the document.

It is noted that new syntax elements are proposed to be in the middle of the structure, so that the decoder is not able to skip them. The syntax elements would need to be moved to the extension.

It was suggested to add as a eoi_object_based_idc bit mask value, but to not include the syntax elements.

Agreed for the eoi_object_based_id mask approach, subject to review of the text that moves the syntax elements to the extension.

Further discussed 10 Oct. 2025.

The text has been confirmed, except that the if condition in the next syntax elements should be compared to 0x20.

Digitally signed content messages (6)

Contributions in this area were discussed during 1715–1730 on Monday 6 Oct. 2025 and 1755 – 2040 Tuesday 7 Oct. 2025 and 1400 – 1430 Thursday 9 Oct. 2025 (chaired by J. Boyce).

General discussion.

Design principles for subpicture:

• Each verification substream and the substreams that it is dependent on, if extracted, would be decodable
• It should be possible to verify that an entire picture is complete, whether or not the picture contains subpictures
• Optionally sets of one or more subpictures may also additionally be able to be verified
  • Options:
    • all subpictures in a set signed together
    • separately signal the set association with the subpictures
• Question: Can an individual subpicture be included in more than one set of subpictures?
• Question: Is it necessary to include parameter sets in the signing of the subpictures/sets of subpictures?

Contribution proposes using a scalable nesting SEI message to associate subpictures.

It was noted that it was proposed in JVET-AM0164 to use scalable nesting SEI for the DSCV SEIs without requiring additional syntax changes.

JVET-AN0208 AHG8/AHG9: Depth-aware optimization for Encoder optimization information SEI message

Scalability dimension info (SDI) SEI