In JCTVC-W0076, an issue was identified with a modification to the pred_weight_table( ) for the "Current Picture Reference" tool (aka IBC), therefore introduced during the development of the SCC extension.

We believe that, while the chosen text solution (layer ID check) is OK, it underlines a lack of clarity, if not an "under-specification", in SHVC. While this was informally discussed with Qualcomm experts, we also had a request for filing this very ticket, so that as many eyes as possible weigh on the issue.

Here's the excerpt we think shows the issue:

• If equalPictureSizeAndOffsetFlag is equal to 1, BitDepthRefLayerY is equal to BitDepthCurrY, BitDepthRefLayerC is equal to BitDepthCurrC, SubWidthRefLayerC is equal to SubWidthCurrC, SubHeightRefLayerC is equal to SubHeightCurrC, and currColourMappingEnableFlag is equal to 0,
  • ilRefPic is set equal to rlPic.
• Otherwise, the following steps apply:
  • The inter-layer reference picture ilRefPic is generated as follows:
  • The PicOrderCntVal value of ilRefPic is set equal to the PicOrderCntVal value of rlPic.
  • The nuh_layer_id value of ilRefPic is set equal to rLId
  • The variable currLayerId is set equal to the value of nuh_layer_id of the current picture.
  • When VpsInterLayerSamplePredictionEnabled[ LayerIdxInVps[ currLayerId ] ][ LayerIdxInVps[ rLId ] ] is equal to 1, the following steps apply:
  • [...]

Please note that:

• The text in italic is our addition and not present in the current text; this is not a proposal, but an example of what we thing could be expected.
• The text in bold hints at why we think it should be there: the reader, and the implementer, would expect explicit setting of values here. But the only reference to layer IDs are through local variables.

In discussing this with Qualcomm's SCC and SHVC text editors, the rationale to consider the current text as sufficient is that the layer setting would be part of a DPB operation.

We are happy to discuss the issue here, which looks to be more SHVC-related than SCC. We consider it an editorial issue, but are unsure of the corrective aspect, if any, and its consequence.