JVET-M0901 BoG report on quantization control related contributions [Y. Ye]

This report was reviewed in Track A on Monday 16 January 1645–1800 (chaired by GJS).

There were 8 technical contributions in the quantization control category.

The first BoG session was held 2000–2330 on 15 Janurary 2019 in Olympia. The second BoG session was held 1300–1510 on 16 January 2019 in Saba.

Section 1 of the document summarizes the BoG’s recommendations. Section 2 of the document contains detailed notes on the BoG discussions. The following aspects were discussed in the track review:

• JVET-M0188 (On quantization parameter signalling considering CU area) and JVET-M0113 (CE7-related: Quantization Group size uniformity)
  • These two proposals solved the same problem of incoherent quantization group (QG) definition in VVC draft 3. They proposed essentially the same solution. However, the two proposals had different ways of expressing their solution in terms of WD text change. The proponents were encouraged to harmonize the suggested WD changes, and the final decision on WD text change should be deferred to the editors. The difference between the two descriptions is purely editorial. Further work was later done to harmonize the text. Decision: Adopt (text in JVET-M0113).
• JVET-M0119 CE7-related: Modified dequantization scaling
  • The second aspect of this proposal was recommended for adoption
  • In VVC draft 3, a scaling factor is applied when a transform block is of rectangular shape. This scaling factor should not be applied in the transform skip case. It was reported that in the reference software VTM-3.0, the scaling factor is applied in the inverse quantization stage and then removes it during the inverse transform stage. However, in VVC draft 3, the scaling factor is applied but not removed, which causes a mismatch between the software and the spec. This contribution removes the (unnecessary) application and removal of this scaling factor for transform skip blocks in the reference software, and proposes a WD change that fixes the mismatch between software and spec. Decision (bug fix): Adopt.
• JVET-M0685 Non-CE7: On derivation of quantization parameter predictor
  • This discusses prediction of the QP value.
  • HEVC had a different QP prediction operation for when WPP is on and off. When WPP is off, any QP change will propagate to all subsequently coded blocks in the slice.
  • In the current draft the QP is predicted as the average of the QP above and to the left. If either of those is outside the CTU, the previous one in decoding order is used as the predictor.
  • This is a problem for parallel encoding, e.g., when encoding a CTU at the left edge, the QP predictor (which is from the right edge) may not be known.
  • JVET-M0685 proposed storing the QP values of the bottom of the above CTU and considering those “available”. If both are available, an average would be used; if one is unavailable, the other would be used.
  • In the discussion of the contribution, another method was discussed, which was to only have the QP of the above CTU available if the current CTU is the first CTU of the row.
  • It was noted that the QP values of the neighbour are already stored for deblocking purposes, so they could be used. Another person commented that the storage for deblocking may be in a different part of the decoding pipeline that would not be readily available.
  • The left-edge-only idea was suggested to potentially be beneficial for rate control in avoiding “dilution” of a delta QP.
  • Decision: Initialize QP from the bottom left CU of the above CTU row when decoding the first CU of a CTU on the left edge of a tile (text was provided in a revision of JVET-M0685).

The BoG recommended further study in an AHG on quantization control. This was agreed to be established.

It was noted that MTS affects the concept of quantization matrices.

JVET-M0901 BoG report on quantization related contributions [Y. Ye]

See section 6.15.