Opened 9 years ago Closed 4 years ago #1449 closed defect (fixed)Incorrect offset derivation in Eq 8-10x from chroma format for CPR
Description
Current formulas in 8.5.3.2.1 are:
These formulae should probably read something like:
Note: there does not seem to be an appropriate version for the SCC text. Change History (3)comment:1 Changed 9 years ago by rajan_joshicomment:2 Changed 4 years ago by crosewarne
Regarding "As a side note, chroma motion vectors are derived at the end of clause 8.5.3.2.1. But they are used in the same clause before derivation. This should be fixed.": This was corrected in JCTVC-W1005. comment:3 Changed 4 years ago by crosewarne
Note: See
TracTickets for help on using
tickets. | This list contains all users that will be notified about changes made to this ticket. These roles will be notified: Reporter, Owner, Subscriber, Participant
|
In clause 8.5.3.2.10, chroma motion vectors are derived as follows:
mvCLX[0] = mvLX[0] * 2 / SubWidthC
mvCLX[1] = mvLX[1] * 2 / SubHeightC
mvLX[0] and mvLX[1] are stored in 1/4 th precision. Since for CPR, the luma motion vector is integer pel, mvLX[0] and mvLX[1] are multiples of 4. Since SubWidthC=1 and SubHeightC=1 for 4:4:4 chroma format, mvCLX[0] and mvCLX[1] are multiples of 8 for CPR. Thus offsetX and offsetY will be 0 for 4:4:4 chroma format.
As a side note, chroma motion vectors are derived at the end of clause 8.5.3.2.1. But they are used in the same clause before derivation. This should be fixed.