#1606 closed defect (fixed)

VVC v2 wrongly uses zero out block size

Reported by: bbross Owned by:
Priority: critical Milestone:
Component: spec Version:
Keywords: Cc: ksuehring, bbross, XiangLi, fbossen, jvet@…

Description

In VVC v2 spec text the zero out transform block sizes Log2ZoTbWidth and Log2ZoTbHeight are used in “9.3.4.2.4 Derivation process of ctxInc for the syntax elements last_sig_coeff_x_prefix and last_sig_coeff_y_prefix” although it was decided by the adoption of JVET-N0194 to use the full block sizes Log2TbWidth and Log2TbHeight (instead of Log2ZoTbWidth and Log2ZoTbHeight).

This bug has been introduced in v2 and should be fixed as follows:

  • In residual_coding() syntax table, move the derivation of Log2FullTbWidth = log2TbWidth and Log2FullTbHeight = log2TbHeight right after Log2ZoTbHeight = Min( log2TbHeight, 5 ), i.e. before last_sig_coeff_x_prefix.
  • In “9.3.4.2.4 Derivation process of ctxInc for the syntax elements last_sig_coeff_x_prefix and last_sig_coeff_y_prefix” replace "Log2ZoTbHeight/Width" with "Log2FullTbHeight/Width".

Change history (1)

comment:1 Changed 10 months ago by bbross

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

To be fixed in both the ITU-T and ISO texts for VVC v3.

Note: See TracTickets for help on using tickets.