Opened 5 years ago

Closed 4 years ago

#613 closed defect (fixed)

Issue with one input to 8.5.6.4 Prediction refinement with optical flow process

Reported by: vdrugeon Owned by:
Priority: minor Milestone:
Component: spec Version: VVC D7 vE
Keywords: Cc: ksuehring, bbross, XiangLi, fbossen, jvet@…

Description

In 8.5.6.1 General, when invoking the prediction refinement with optical flow process, the motion vector difference array diffMvLX[ xIdx ][ yIdx ] is given as input. However, xIdx and yIdx are not defined at this point. Another issue is that if only one value of the array is given as input, this is not an array anymore, but 8.5.6.4 Prediction refinement with optical flow process takes an array as input.
Although the way this array is used in 8.5.6.4 is strange, because it seems to use pixel positions as index, whereas the array diffMvLX as defined in 8.5.6.1 seems to rather take sub-block positions as index.
Please check!

Change history (5)

comment:1 Changed 4 years ago by bbross

  • Version set to VVC D7 vC

comment:2 Changed 4 years ago by bbross

  • Version changed from VVC D7 vC to VVC D7 vD

comment:3 Changed 4 years ago by bbross

  • Version changed from VVC D7 vD to VVC D7 vE

comment:4 Changed 4 years ago by jlchen

diffMvLX[ xIdx ][ yIdx ] always take luma sample location as input. The input (diffMvLX[ xIdx ][ yIdx ] ) of 8.5.6.4 (invoked in 8.5.6.1) should be an array with xIdx = 0..cbWidth / numSbX − 1, and yIdx = 0..cbHeight / numSbY – 1

comment:5 Changed 4 years ago by jlchen

  • Resolution set to fixed
  • Status changed from new to closed
Note: See TracTickets for help on using tickets.