The HM decoder crashes when field_seq_flag is equal to 1 due to the following code line in file TDecCAVLC.cpp

assert(pcVUI->getFieldSeqFlag() == false);

After removing the above code, the decoder works fine by outputing field pictures sequentially.

Michael

HM decoder crashes when there is no payload_bit_equal_to_zero at the end of a SEI message. Specifically, the following code

READ_CODE (payloadBitsRemaining-1, dummy, "payload_bit_equal_to_zero");

will trigger an assert failure when payloadBitsRemaining is equal to 1

assert ( uiLength > 0 );

A patch is attached to fix this issue.

Best regards, Michael

There are three redundant cbf_cb and cbf_cr flags for each 4x4 chroma TU when log2MaxTrafoSize is set to two.

When log2MaxTrafoSize is set to two, there are only 4x4 luma TUs and 4x4 chroma TUs and every four luma TUs correspond to one cb TU and one cr TU. So only one set of cbf_cb and cbf_cr flags should be coded corresponding to every four cbf_luma flags.

However, HM 5.1 software still codes one cbf_cb and cbf_cr flags corresponding to each cbf_luma when log2MaxTrafoSize is set to two, which causes three redundant cbf_cb and cbf_cr flags for each 4x4 chroma TU.

The relevant code causing this behavior is in TDecEntropy::xDecodeTransformSubdiv

Bool bFirstCbfOfCU = uiLog2TrafoSize == pcCU->getSlice()->getSPS()->getQuadtreeTULog2MaxSize() || uiTrDepth == 0

It does not consider the case when log2MaxTrafoSize is set to two. So for each 4x4 luma TU, bFirstCbfOfCU is always true and there will always be a set of cbf_cb and cbf_cr coded.