HTM15.1 and later decoders crash at POC18/Layer1 in 2-view + 2-depth bit stream of 3D-HEVC mode. HTM15.0 decoder doesn't.

In Multiview Main profile, there is the following conformance restiction. "– When NumLayersInIdList[ OlsIdxToLsIdx[ olsIdx ] ] is equal to 2, output_layer_flag[ olsIdx ][ j ] derived according to any active VPS shall be equal to 1 for j in the range of 0 to 1, inclusive, for subBitstream."

In HTM encoding with the current cfg files, the restriction fails:

Solving this issue, we should remove the following two lines in cfg/MV-HEVC/baseCfg_2view.cfg OutputLayerSetIdx : 1 # Indices of layer sets used to derive additional output layer sets LayerIdsInAddOutputLayerSet_0 : 1 # Indices in VPS of output layers in additional output layer set 0

And we should replace the following lines in cfg/MV-HEVC/baseCfg_3view.cfg OutputLayerSetIdx : 1 2 # Indices of layer sets used to derive additional output layer sets LayerIdsInAddOutputLayerSet_0 : 1 # Indices in VPS of output layers in additional output layer set 0 LayerIdsInAddOutputLayerSet_1 : 2 # Indices in VPS of output layers in additional output layer set 1 with OutputLayerSetIdx : 2 # Indices of layer sets used to derive additional output layer sets LayerIdsInAddOutputLayerSet_1 : 2 # Indices in VPS of output layers in additional output layer set 1

Note: For 2 view case, I've conducted simulation with cfg change and confirmed the result but for 3 view case, I haven't conducted the simulation.

In addition, we may have the checks in decoder/encoder.

Decoder sides's check code may be as follows:

Void TDecCavlc::parseVPSExtension( TComVPS* pcVPS ) {

... for( Int i = 1; i < pcVPS->getNumOutputLayerSets( ); i++ ) {

if( pcVPS->getNumLayerSets() > 2 && i >= pcVPS->getNumLayerSets( ) ) {

READ_CODE( pcVPS->getLayerSetIdxForOlsMinus1Len( i ), uiCode, "layer_set_idx_for_ols_minus1[i]" ); pcVPS->setLayerSetIdxForOlsMinus1( i, uiCode );

} ... if( pcVPS->getNumOutputLayersInOutputLayerSet( i ) == 1 && pcVPS->getNumDirectRefLayers( pcVPS->getOlsHighestOutputLayerId( i ) ) > 0 ) {

READ_FLAG( uiCode, "alt_output_layer_flag[ i ]" ); pcVPS->setAltOutputLayerFlag( i, uiCode == 1 );

}

#if NH_MV_FIX_TICKET_XXX

if (pcVPS->getNumLayersInIdList( pcVPS->olsIdxToLsIdx( i ) ) == 2 && pcVPS->getPTL(pcVPS->getProfileTierLevelIdx(i, 1))->getGeneralPTL()->getProfileIdc() == Profile::MULTIVIEWMAIN) {

for( Int j = 0; j < pcVPS->getNumLayersInIdList( pcVPS->olsIdxToLsIdx( i ) ); j++ ) {

if (pcVPS->getOutputLayerFlag(i, j) == 0) {

printf("!!Warning!!, output_layer_flag[] shall be always 1 when two layer is used in OLS in MV-HEVC Main profile\n");

}

assert(pcVPS->getOutputLayerFlag(i, j));

}

}

#endif

}

About a 3D-HEVC cfg file, tags/HTM-15.1/cfg/3D-HEVC/baseCfg_2view+depth.cfg:

To use MPI as default in 2 view case, we should replace the value: Log2MpiSubPbSizeMinus3 : 3 with Log2MpiSubPbSizeMinus3 : 0

Note: 3 view case(baseCfg_3view+depth.cfg) or other cfg files seems correct.