JVET-R0047 AHG9/AHG12: On slice address for raster scan slices in a picture [L. Chen, C.-W. Hsu, Y.-W. Huang, S.-M. Lei (MediaTek)]
This contribution proposes an optimization to derive slice address for raster scan slices in a picture.
In V2 of this contribution, the software is attached.
For raster scan slices, we signal both a slice address and the number of tiles in the slice (minus 1). The proposal is to infer the slice address from the ending position of the previous slice in decoding order.
Conceptually, with this approach, we would not need to have had slice addresses sent in AVC (if the decoder would parse the previous slice to figure out how many macroblocks were in it).
It was commented that this would remove the ability to decode a slice independently, and would thus have packet loss implications.
We currently have a syntax and decoding process such that any slice can be decoded separately and independently of all other slices in the picture – a property shared by AVC and HEVC (and probably MPEG-2 and H.263). This proposal would remove that property.
It was commented that in some systems, packets can arrive out of order. The proponent said this can be handled at a different layer.
No action was taken on this.