Opened 5 years ago

Closed 5 years ago

#323 closed defect (fixed)

Mismatch in Affine MVP candidate list construction between VTM and spec

Reported by: chhuanb Owned by:
Priority: minor Milestone: VVC D6 vD
Component: spec Version: VVC D6 vC
Keywords: Cc: ksuehring, bbross, XiangLi, fbossen, jvet@…

Description

In spec, clause 8.5.5.7, step 7, the inserted order of spatial candidates is 0..2 (above-left, above-right, left-bottom).
In SW, function fillAffineMvpCand, the order is 2..0 (left-bottom, above-right, above-left).

Change history (8)

comment:1 Changed 5 years ago by hanhuang

The spec should be modified to align with SW. The related proposal is JVET-L0271. The provided text in JVET-L0271 is as following:

When numCpMvpCandLX is less than 2, for cpIdx range from 2 to 0 and numCpMvpCandLX less than 2 the following applies:
– If availableFlagLX[ cpIdx ] is equal to 1, the following assignments are made:
– cpMvpListLX[ numCpMvpCandLX ][ 0 ] = cpMvpLX[ cpIdx ] (8 xxx)
– cpMvpListLX[ numCpMvpCandLX ][ 1 ] = cpMvpLX[ cpIdx ] (8 xxx)
– cpMvpListLX[ numCpMvpCandLX ][ 2 ] = cpMvpLX[ cpIdx ] (8 xxx)
– numCpMvpCandLX = numCpMvpCandLX +1 (8 xxx)

comment:2 Changed 5 years ago by bbross

  • Version changed from VVC D5 v8 to VVC D5 v9

comment:3 Changed 5 years ago by bbross

  • Version changed from VVC D5 v9 to VVC D5 v10

comment:4 follow-up: Changed 5 years ago by bbross

Does this issue still exists in D6 vC?

comment:5 in reply to: ↑ 4 Changed 5 years ago by chhuanb

Replying to bbross:

Does this issue still exists in D6 vC?

Yes. step 7 should be modified as:

  1. The following applies for cpIdx = 2..0:

comment:6 Changed 5 years ago by bbross

  • Milestone set to VVC D6 vD

comment:7 Changed 5 years ago by bbross

  • Version changed from VVC D5 v10 to VVC D6 vC

comment:8 Changed 5 years ago by bbross

  • Resolution set to fixed
  • Status changed from new to closed

Will be fixed in vD

Note: See TracTickets for help on using tickets.