In SHM-5.1-dev r614, the DPB-parameters in SPS are inferred when they are not signalled for nuh_layer_id greater than 0. The following inference in TDecCavlc.cpp:parseSPS() is incorrect because the inference rule is dependent on which layer accesses the SPS; that information will be unavailable at the point of parsing the SPS and hence the following should not be included in parseSPS().

for(UInt i=0; i <= pcSPS->getMaxTLayers()-1; i++) {

pcSPS->setMaxDecPicBuffering( parameterSetManager->getPrefetchedVPS(pcSPS->getVPSId())->getMaxVpsDecPicBufferingMinus1( parameterSetManager->getPrefetchedVPS(pcSPS->getVPSId())->getTolsIdx(), pcSPS->getLayerId(), i)+1, i);

}

In some places, the SPS layer ID is used instead of the slice layer ID, which could be wrong in many cases.

Attached ticket fixes this issue.

When bit-depth scalability is enabled with 1x spatial scalability, SHM-5.0 software does not match the specification of the resampling process in section H.8.1.4 of SHVC working draft 5.

SHVC draft 5 section H.8.1.4 specifies that when the base layer and enhancement layer have the same picture sizes, same internal bit-depths, and zero scaled reference layer offsets, the inter-layer reference picture ilRefPic is set equal to the decoded reference layer picture rlPic.

However in SHM-5.0, when the base layer and enhancement layer have the same picture sizes and zero scaled reference layer offset, the inter-layer reference picture ilRefPic is set equal to the decoded reference layer picture rlPic without checking the bit-depth difference of BL and EL. Consequently, SHM-5.0 does not support bit-depth scalability with 1x scalability in the inter-layer prediction process.

This bug was reported in JCTVC-Q0130, and two implementations of the bug fix are in the software attached in JCTVC-P0130-v2 with the following two macros:

• BIT_DEPTH_AND_1X_SCALABILTY_FIX1
• BIT_DEPTH_AND_1X_SCALABILTY_FIX2