Undefined variable “maxTbSize” appears in the following subclauses:

8.4.5.1 General decoding process for intra blocks 8.5.8 Decoding process for the residual signal of coding blocks coded in inter prediction mode 8.5.9 Decoding process for the reconstructed signal of chroma coding blocks coded in inter prediction mode

Probably the text "Depending on maxTbSize, the following applies:" should be

"Depending on maxTbWidth and maxTbHeight, the following applies:"

(This seems better than using 'MaxTbSizeY' and simpler than trying to make more significant changes like listing all the inputs in the preamble to the following section, or identifying the output rather than the input(s) to the section in this preamble)

Version VTM-17.0

After sh_no_output_of_prior_pics_flag is parsed

READ_FLAG(uiCode, "sh_no_output_of_prior_pics_flag");   pcSlice->setNoOutputOfPriorPicsFlag(uiCode != 0);

it is soon reset like this:

    //the inference for NoOutputOfPriorPicsFlag
    if( !m_firstSliceInBitstream && m_picHeader.getNoOutputBeforeRecoveryFlag() )
    {
      m_apcSlicePilot->setNoOutputOfPriorPicsFlag(true);
    }
    else
    {
      m_apcSlicePilot->setNoOutputOfPriorPicsFlag(false);
    }

where NoOutputBeforeRecoveryFlag is set to 1 for every IDR slice.

So it doesn't matter what value was signaled for sh_no_output_of_prior_pics_flag.

This cause lost frames in Multilayer streams when dpb_max_dec_pic_buffering and dpb_max_num_reorder_pics are greater than necessary.

Consider 2 layers, both AllIntra

Layer0: LId0-POC0(IDR) LId0-POC1(IDR) Layer1: LId1-POC0(IDR) LId1-POC1(IDR)

dpb_max_dec_pic_buffering = dpb_max_num_reorder_pics = 5 (or anything >1)

Pictures encoded in following order:

• LId0-POC0 (IDR)
• LId1-POC0 (IDR)
• LId0-POC1 (IDR)
• LId1-POC1 (IDR)

When decoder starts decoding picture LId0-POC1 it will output LId0-POC0 via xFlushOutput since LId0-POC1 is an IDR. But not LId1-POC0, since it is from different layer.

Then decoder derives NoOutputOfPriorPicsFlag = true and calls checkNoOutputPriorPics() which sets neededForOutput=false for all frames in DPB.

As a result, LId1-POC0 is never outputted.

In “8.4.3 Derivation process for chroma intra prediction mode": If treeType is equal to SINGLE_TREE, sps_chroma_format_idc is equal to 3, intra_chroma_pred_mode is equal to 4, and IntraMipFlag[ xCb ][ yCb ] is equal to 1, the following applies: – The MIP chroma direct mode flag MipChromaDirectFlag[ xCb ][ yCb ] is set equal to 1.

When cu_act_enabled_flag is equal to 1, intra_chroma_pred_mode is not present in bitstream, and according to specification, "When intra_chroma_pred_mode is not present, it is inferred to be equal to 0.". As the result, MipChromaDirectFlag should be 0.

However, according to CABACReader::intra_chroma_pred_mode() of CABACReader.cpp: if (pu.cu->colorTransform)

{

pu.intraDir[CHANNEL_TYPE_CHROMA] = DM_CHROMA_IDX; return;

}

and PU::isMIP() in UnitTools.cpp considered "(pu.intraDir[CHANNEL_TYPE_CHROMA] == DM_CHROMA_IDX)" as one of the condition to judge MipChromaDirectFlag.

As the result, under the conditions below:

• treeType is equal to SINGLE_TREE &&
• sps_chroma_format_idc is equal to 3 &&
• IntraMipFlag[ xCb ][ yCb ] is equal to 1 &&
• cu_act_enabled_flag is equal to 1

specification & VTM will give different result of MipChromaDirectFlag, due to different default value for intra_chroma_pred_mode when cu_act_enabled_flag is 1.