There is a mistake in the output reconstruction file in the SHM-8.0.

In the output reconstruction file process the output bitdepth is not initialized.

Please find in attachment a patch for TappDecTop.cpp(It is generated with the SHM8.0 rev 1026).

It is necessary to do more test. The patch corrects only one method. The output reconstruction file is generated in several places in the code.

According to "F.7.3.2.3.3 Picture parameter set multilayer extension syntax", the index for scaled_ref_layer_offset_present_flag[i], ref_region_offset_present_flag[i] and resample_phase_set_present_flag[i] is based on num_ref_loc_offsets. [ref: for( i = 0; i < num_ref_loc_offsets; i++ )].

However, in TEncTop.cpp, TEncTop::xInitPPS()

for(Int i = 0; i < m_cPPS.getNumScaledRefLayerOffsets(); i++) {

#if O0098_SCALED_REF_LAYER_ID

m_cPPS.setScaledRefLayerId(i, m_scaledRefLayerId[i]);

#endif

m_cPPS.getScaledRefLayerWindow(i) = m_scaledRefLayerWindow[i];

#if REF_REGION_OFFSET

m_cPPS.getRefLayerWindow(i) = m_refLayerWindow[i]; m_cPPS.setScaledRefLayerOffsetPresentFlag( m_scaledRefLayerId[i], m_scaledRefLayerOffsetPresentFlag[i] ); m_cPPS.setRefRegionOffsetPresentFlag( m_scaledRefLayerId[i], m_refRegionOffsetPresentFlag[i] );

#endif #if R0209_GENERIC_PHASE

m_cPPS.setResamplePhaseSetPresentFlag( m_scaledRefLayerId[i], m_resamplePhaseSetPresentFlag[i] );

The index used is based on m_scaledRefLayerId[i]. This one should just be 'i'.

Where the structure is referenced, it uses an index based on num_ref_loc_offsets.

for(Int i = 0; i < pcPPS->getNumScaledRefLayerOffsets(); i++) {

WRITE_CODE( pcPPS->getScaledRefLayerId(i), 6, "ref_loc_offset_layer_id" ); WRITE_FLAG( pcPPS->getScaledRefLayerOffsetPresentFlag(i) ? 1 : 0, "scaled_ref_layer_offset_prsent_flag" );

... ...

WRITE_FLAG( pcPPS->getRefRegionOffsetPresentFlag(i) ? 1 : 0, "ref_region_offset_prsent_flag" ); if (pcPPS->getRefRegionOffsetPresentFlag(i))

Please see the patch in the attachment. This is based on rev 925 latest dev branch. This one will affect only where there are more than 2 layers.

SHM software drifts from the Spec for some VPS conditions. Please look at the patch for the fixes. The patch is on latest SHM rev 953.

I have encapsulated all the fixes under the #define VPS_FIX_TO_MATCH_SPEC so as to be able to compare with and without the fixes.