As we all know, to remove the dependency of deriving the spatial candidate of upper position on left position, the simplification 2 of JCTVC-G542 defines a new derivation way of the parameter isScaledFlagLX and be adopted in the 7th meeting.

However, in the function "fillMvpCand" of the file "TComDataCU.cpp" of HM6.0-rc1, before the scaling calculation for upper position, there are two statements:

bAdded = bAddedSmvp; if (pInfo->iN==2) bAdded = true;

In my opinion, the 2nd statement can be deleted, as we only can derive the value of pInfo->iN after checking all the left positions. As a result, we did not really eliminate the dependencies between derivation of the spatial candidate of upper position and left position.

After we carry out this modification, the experiment results is not affected, as the 2nd statement has no substantive role, whereas we really eliminate the dependency.

When we did expeiments on HM4.0, we observed that the following section of the code of the function "encodeCoeff "in the file "TEncEntropy.cpp" does not seem to behave in the correct way.

if ( !pcCU->getQtRootCbf( uiAbsPartIdx ) )

{

return;

}

It seems that the above code should be modified as follows.

if ( !pcCU->getQtRootCbf( uiAbsPartIdx ) )

{

pcCU->setCbfSubParts( 0, 0, 0, uiAbsPartIdx, uiDepth ); pcCU->setTrIdxSubParts( 0 , uiAbsPartIdx, uiDepth ); return;

}

In this way, it can also match the following section of the code of the function "decodeCoeff" in the file "TDecEntropy.cpp"

if ( !uiQtRootCbf )

{

pcCU->setCbfSubParts( 0, 0, 0, uiAbsPartIdx, uiDepth ); pcCU->setTrIdxSubParts( 0 , uiAbsPartIdx, uiDepth ); return;

}

The following section of the code of the function "encodeCoeff " in the file "TEncEntropy.cpp" of HM6.0 also has the same problem. if ( !pcCU->getQtRootCbf( uiAbsPartIdx ) )

{

#if NSQT_LFFIX

pcCU->setNSQTIdxSubParts( uiAbsPartIdx, uiDepth );

#endif

return;

}

It seems that it should make the following changes. if ( !pcCU->getQtRootCbf( uiAbsPartIdx ) )

{ pcCU->setCbfSubParts( 0, 0, 0, uiAbsPartIdx, uiDepth ); pcCU->setTrIdxSubParts( 0 , uiAbsPartIdx, uiDepth );

#if NSQT_LFFIX

pcCU->setNSQTIdxSubParts( uiAbsPartIdx, uiDepth );

#endif

return;

}

for( cIdx = 0; cIdx < (chroma_format_idc == 0) ? 1 : 3; cIdx++ )

would result in an infinite loop... substitute with this:

for( cIdx = 0; cIdx < (chroma_format_idc == 0 ? 1 : 3); cIdx++ )

This error is repeated in the pseudocode in "Decoded picture hash SEI message semantics"